'OK, everyone. Stop typing, this software is DONE,' said no one ever I recently received an email from an old shared hosting provider. The host wanted to "upgrade" my account to a new server. I had long since stopped using the account for all but one client site that ran a legacy version of Django. I built it ages ago, but it was done. It worked fine and the client was still happy with it. No big …
Your plumbing analogy is shit.
A better one would be that your appliances that you bought to run on a crazy old pre-standardisation 170v @ 40hz electricity supply is getting cut off, you need to buy new shit ( or bodge the old shit with transformers and magical frequency adjustifiers [i'm not a spark, did you guess?]) to get it to work.
If you want the 170v @ 40hz supply then fine, but you're going to have to pay through the nose for it.
Agreed, poor analogy.
You don't rent the shower off the plumber. Nor does the plumber operate it for you, clean it and ensure that it has electricity and water supplied to it.
A better analogy would be:
If you paid me to use my shower because you don't want to have one in your house and I choose to relocate the shower to my kitchen because it makes it more convenient for me. That's a reasonable choice and you just have to live with it or find another shower.
Ok, here goes with my analogy.
I have seven rabbits, six ducks and a pelican.
I put the pelican in a nature reserve for pelicans so it's safe and happy.
One day someone comes along and changes the reserve and the pelican is not happy, in fact I have to go and console the pelican just to make it happy once more.
The ducks are upset I've not made any bread and the rabbits don't get the carrot.
The moral of the story is that you never ever mark code as done because there will always be changes.
Actually it's lower case v for Volt because it's alternating current and not constant current.
Upvote for humorous attempt at trolling IT beginners. Nice, subtle, and supported by a plausible sounding argument. That's an 8/10 from me (7/10 plus brevity bonus).
I agree. I lost the count of plumbers who forced me to "upgrade" something because spare parts were no longer available (sometimes lying blatantly, but there was a time when it was difficult to check), or some new policy required to get rid of old stuff when replaced. Sometimes "upgrading" also meant breaking tiles, and other masonry work...
A better one would be that your appliances that you bought to run on a crazy old pre-standardisation 170v @ 40hz electricity supply is getting cut off, you need to buy new shit
More like you built a town to run on the standard 230V 50Hz, then one day the electricity board decided to put in new generators, and switched to 120v 60-Hz because their new designer was American and he felt that obviously 120v was superior.
The true problem isn't that something is "done", it's that too many opensource designers don't give a flying fig about compatibility. Why shouldn't something written for CentOS 6 not still run on CentOS 7? I have stuff compiled 25 years ago on Solaris 2.2 on a SPARCstation 2 that still runs just fine on Solaris 11.3 on a SPARC T5 today.
"It causes no end of fun when people and businesses move, or "consultants" forget to check the supply at the actual site."
A new mainframe was delivered to a UK customer. Due to long UK production delays it was decided to import a USA market version. The customer was very pleased it had finally arrived.
The supplier's engineers commissioned it OK with a 110V transformer lashed into the 240V mains supply. They then left the customer's electricians to integrate the transformer into their infrastructure.
After switch on the next day it was discovered that the customer's electricians had wired the transformer the wrong way round.
Firefox. I miss the old reliable with low memory use, fast, responsive and looked like firefox. Still use it but still miss the older versions.
Have you noticed that since they put multithreading in to make it "faster," that when you open the javascript console, you've got 30 seconds before it populates itself.
That's thirty. Seconds. It's pretty astonishing.
@Bucky 2 - "Have you noticed that since they put multithreading in to make it "faster," that when you open the javascript console, you've got 30 seconds before it populates itself."
If you're talking about the "Browser Console" (there's nothing specifically named "Javascript Console"), it populates on change. Do something which causes Javascript to execute, and you'll get messages, provided you have warnings and errors enabled.
I just did a "preview" on this post, and I get a series of Javascript warnings immediately (use of deprecated features, things which will slow down page rendering, etc.).
I'm using Firefox 54 on Ubuntu.
like me, a quick browse through the Screwfix catalogue will reveal that the hammer isn't quite done yet. All manner of materials creep into the design, all manner of refinements and enhancements... precision shot weighted dampening, hickory tang inserts, 3D printed head rivet pins, magnetic holders, precision ground anti-slip striking faces, anti-sweat bio-active grip windings...
Of course taking a shoe off and whacking in a pin to hold up that new print from the little corner shop that sells dust collecting motivational framed pictures is mostly good enough for some people.
This post has been deleted by its author
Screwfix for tool porn ??? You amateur !!!
The guys who just decorated the outside of our place in Branscombe alerted me to this outfit, where they spend all their spare cash. And they have some awesome tools. I never knew a rail saw existed before, but it's absolute magic. Festool are the canine gonads, apparently.
https://www.axminster.co.uk/
Damn you AC. That looks like some good stuff. (I like their selection of Engineering lathes and mills. Seems to be the standard "straight from China in a crate" design but the fit and finish looks a lot better. The price doesn't but that would be justified if the quality meets the looks.
Isn't it just B&Q with the prices jacked up a bit. Well, at least that's how is seems here and the two are less than 400yds apart in Farnborough.
I agree about axminster though. Enough serious kits to make a tool fetishist go weak at the knees but searching for "40tpi taps" seems to return only saws.
@Steve Davies 3
Screwfix and B&Q are parts of Kingfisher and have a lot in common. Buy your cheap and cheerful tools at B&Q but get all your parts from Screwfix where a packet of 100 costs much the same as a pack of 10 at B&Q. The downside is that if you do a lot of varied building work your shed fills up with the various packs of 85 bits from Screwfix that you didn't use leaving no room for the sexy power tools you can get from a serious tool supplier.
festool....buy one. no wait, dont buy one. you will end up replacing all of your tools and have no money.
they are the doggies bollox. the festool tracksaw i bought is fecking unbelievable.
i almost bougth 400 quid router (yes thats router not router) that i didnt need. i have used my existing router twice in 10 years. but i needed a new router according to my festool infected brain. i only just managed to resist
Not so good your hammer analogy, they are still all hammers. The gimmicks/and tweaks are largely cosmetic and don't change the basic bashing functionality expected.
These features are of course the one your boss will get excited about saying he needs to spend on the most expensive hammer with all the features (that will never be used in reality) just in case after having a beer with the salesman.
Drywall and Shingling hammers are half axe. A slide hammer doesn't look anything like a hammer. Veneering hammers look like hammers but you don't use them to belt the veneer into place.
I'm appalled at the lack of basic tool knowledge by so-called "engineers".
I like hammers. I even wrote a song about them when people kept telling me to go to my "Happy Place" some years ago titled "My Happy Place Has A Hammer In It (BANG-BANG-BANG-BANG-BANG)".
Also a fan and owner of a basement full of power tools including sanders (table and portable, pad and belt), routers (plunge and fixed), saws (chain, reciprocating and circular, table and portable ranging from dangerous to "you must be f*cking joking" in the case of the chainsaw on a ten foot pole and the repurposed angle grinder saw of certain death), drills (floor-standing, portable, non-hammer and hammer - for some cross genre tooly fun), and a rather nice biscuit joiner. Oh, and a couple of Dremel Rotary tools, a Dremel oscillating cutter, a Dremel lathe (naff, but fun) and the Delta Chopsaw of Extreme Spiffiness and Universal Utility, best use of money ever spent.
In the garage there are two weedwackers, one that I can't get spares for (ding for relevance to article) and one which can mount a circular saw on the end (Yeeha!) a power mower, Troll (the Snow Blower of Supreme Spiffiness), a Toro brand leaf blower/vacuum that was once rated as the most powerful in the world and had me doing the helicopter dance the first time I turned it on and two hedgeclippers (corded and cordless). And another chainsaw.
And that's just the stuff with an electric or gasoline motor built-in. I also have an awesome .22 gauge concrete nail bazooka, but that has no continuous mode and must be reloaded every time I want to have some loud nail-based fun.
I forgot the bandsaw I picked up in a garage sale a while back but haven't used yet.
No. Too lame. (Tool lame).
Besides, I don't need one. I have a chainsaw-onna-ten-foot-pole.
I liked that story.
A 100+year old tree outside a supermarket fell down last winter - miraculously falling just short of most of the surrounding buildings. It was decided to pollard the remaining tree of similar enormous girth and height - it couldn't be felled because it is in a conservation area. The main branches were cut at a point where they were at least a foot (30cm) thick. One fell in the wrong place - and reduced a section of substantial 4x4 timber car park fencing to matchwood.
I have been tempted to use a chainsaw on a pole to remove an overhanging branch of next door's feral tree that has grown in a few years to block my sunlight. The neighbour is totally uncooperative - so every tree surgeon has refused to touch the job. There seem to be two potential disasters.
1) the chainsaw can get stuck in the cut at maximum reach.
2) the branch could "hinge" on the final strip of bark just before the cut is complete. The resulting trajectory would then smash it into through the neighbour's fence.
So all I can do is pray that the tree is not immune to the "ash die-back" disease that is apparently ravaging the UK's ash trees.
Not so good your hammer analogy, they are still all hammers. The gimmicks/and tweaks are largely cosmetic and don't change the basic bashing functionality expected.
I guessing you don't know your ball-peen from your dead blow, or your soft face from your lump hammer. Not every hammer is the thing you use for nails.
The big difference is that software doesn't wear out from usage, but real-world items typically do.
Take that hammer, if you hit enough nails with it, eventually the head will likely come loose and the shaft may crack, meaning you have to repair or replace it. Use your office chair enough, and eventually the cover and padding will begin to disintegrate and you'll need to replace it.
A lot of physical items therefore have a guaranteed revenue stream as you'll need to buy new ones when the old ones fail (more and more items are *designed* to fail after a while of course, but this is another story).
Software doesn't do this though. You don't wear out a loop by iterating through it too many times, hence the software writers have to come up with new methods to maintain revenue. Typically you either keep upgrading the software and depreciating old versions to force people to upgrade, or you charge for expensive support contracts (and make sure your software breaks in complicated ways from time to time to justify the support), or of course the latest method is the dreaded "subscription" approach, or annual license.
Maybe with subscription approaches, some software could increasingly be considered as "finished", in that you get revenue regardless of whether you create new versions. The only issue is whether a competitor releases something better and pinches your customers.
From here, we're back to physical products, where improvements are simply meant to ensure people continue buying your company's stuff when their old items break. Like that nice tungsten-tipped hammer with the comfort-grip handle...
If we were in the habit of considering software "done", I'd be typing this on vi and uploading it using FTP to a bulletin board.
Software is never 'done' because it can always be improved upon. No company is going to stop developing their software, sit back, and expect to continue selling the same software for the next 20 years. Their competitors would put them out of business within 5.
The tricky bit is, of course, determining what constitutes an improvement, which can be very much a matter of opinion.
> The big difference is that software doesn't wear out from usage, but real-world items typically do.
The big difference is that real software has security vulnerabilities which are not known at the time it was written, but come to light later and are exploited by bad people.
I hate the analogy, but it's as if the plumber had installed a shower with a remote control, and it turned out later that the remote control also allowed people to open your front door. At that point, you're really forced to fix the shower, or else remove it entirely and use the bath instead.
Hence software is never "done". If a project is abandoned due to lack of interest by the developers, there is risk in continuing to use it - unless you're prepared to fix it yourself.
This is acknowledged by proper software which has built-in obsolescence dates, such as the example given of CentOS 6 going EOL in 2020. It's annoying if your VPS provider turns off support before then.
However Django also has its own separate lifecycle. If your project uses anything other than Django 1.11, 1.10 or 1.8 then you are using unsupported software:
https://www.djangoproject.com/download/#supported-versions
...and hence you are at extremely high risk, given that security problems in web frameworks are a pretty regular occurrence. Leaving your application to rot on a VPS is not a sensible option.
It wouldn't matter. Even if YOU don't use it, someone else COULD and probably WILL.
How about this? Something in the door's design crops up that makes it extremely easy for burglars to jimmy their way in, mandating a change to the design to make it harder to jimmy because a spate of burglaries took advantage of this design flaw.
Are you so sure about hammers being 'done'? Claw, ball-peen, dead-blow, lump, sledge, toffee... Wood, steel, rubber, plastic... How about the handles - again, wood, steel, or maybe plastic (GRP, ABS...) Not to mention power hammers...
I realise that probably wasn't your intention with the analogy and I'm being somewhat pedantic, but hammers, like software, do actually get tweaked and adjusted when what exists doesn't work properly for the current situation.
Now if you were renting a hammer (hammer as a service?) or a shower, you may very well find that the provider does decide at some point that what they're willing to offer does change so you end up in a forced upgrade situation, much like your hosting provider.
"[...] you may very well find that the provider does decide at some point that what they're willing to offer does change so you end up in a forced upgrade situation, [...]"
Reminds me of a "joke" told by an old timer at least 50 years ago.
Person asks the shopkeeper for a replacement for a valued tool. The shopkeeper replies "You must be the 50th person I've told today - there's no market for them any more"
So... yes, CentOS 6 is "supported" until 2020 but the upstream RHEL that it's based on and receives all its patches from went into what Redhat call "Production Phase 3" on May 10th 2017. That means that only security vulnerabilities that Redhat class as "critical" will now be fixed. Anything that's merely "important" or less gets patched if they feel like it and judging by the things that _didn't_ get fixed in RHEL5 once that reached PP3, that's going to mean that CentOS 6 gets less and less secure over the next 2.5 years until it goes EOL.
Also, if the host uses cPanel on their boxes for management etc, and the server was running a 32-bit version of CentOS (possible on 6, not on 7), then they'd need to update sooner since cPanel are ending support for 32-bit after v56. While the EOL has extended to October, it was originally supposed to be April, after which you'd no longer receive any cPanel updates if you didn't move to a 64-bit version.
I think I know which company did this. they greatly inconvenienced their customers (including one of mine).
If the customer wants to STAY WITH WHAT HE HAS, he should be able to.
It's BAD BUSINESS to "force" like that. Especially when it's done WITHOUT WARNING like that. It would be like changing their web server from Apache to nginx (or the other way around) and then it's like "oh, you didn't get the memo? here it is..."
or worse - from Linux to a WINDOWS server - without warning. "Case-sensitive file names? That actually MATTERS?" (among other things)
Well some software does get to done.
Especially embedded software, in items, vehicles, and especially in defence.
There is no reason someone should be fiddling with my car's ECU once it is a product. (ignoring Volkwagon and others doing naughty things with their emissions).
And while at the moment the software in the F35 is under development and there are new releases and changes, there will come a point where it is done, and only gets changed to add new things in the future.
And then it will be 20+ of unchanged.
Contradicted by his own argument:
"In most cases, though, what works now works well enough. Even lowly, much-maligned Windows XP still works well enough. The only reason it's insecure is that its creator decided to stop making it secure."
So quite clearly Windows XP wasn't done otherwise it wouldn't need security fixes.
The argument is also flawed in it ignores all the software products which don't get updated. Mostly small simple tools that do a single job but I'm sure if you look you will find many *nix tools which haven't had any meaningful changes in a long time even though they are regularly used. For everything else the fundamental reason is that requirements change, hardware changes, new ideas come along and so there is always more that could be done to make something 'better'.
Um, I think it could still be done: if it met the requirements document at the time of sign-off. New vulnerabilities are those not considered in that document, no?
(not getting into whether that's a good thing or not, but I think there is a distinction between 'software does what it was designed to do' and 'software has a post-finished requirements change'. The latter is clearly maintenance. My fence was finished years ago, but it now needs painting again.)
When you said, "The site I hosted with that company didn't need an upgrade, it was running fine with almost no attention at all", what you really meant was: "I was not interested in bearing the cost of maintaining my fossilised software, and got a bit grumpy about the whole thing when my service provider made it clear that they were no longer interested in bearing the cost of maintaining my fossilised software either".
Centos 6 dates from 2011 and ships with packages and core o/s software from that era. It is completely true that an admin can stick with these operating system versions in the longer term, and this is a viable option which is regularly used by high value software installations. The problem is that there is a substantial cost in doing so and when you're maintaining low-cost sharing hosting services, costs matter quite a bit.
So it's not just about sticking with centos 6: it's about sticking with e.g. python 2.6, php 5.3, apache2.2 and a pile of other ancient software versions which are no longer supported by their developers and which are basically moribund. It's about committing to your user base that the hosting company will assume responsibility for fixing security problems in Django because the oldest maintained release of Django was released in 2015 and requires python 2.7. It's about not being able to provision any more customers on servers like this because you simply cannot sell python 2.5 and php 5.3 in 2017, which means that according as customers migrate off this platform, you're ending up with a much lower density of customers per physical server, which drives up the cost.
It's about the tool chains and automation software (e.g. puppet / ansible / salt) required to support cabinets full of older servers like that, which drives up the cost.
It's about having a support desk that can deal with issues relating to lots and lots of different software versions. It's about installing multiple software versions on these platforms and attempting to ensure that none of them are fighting with each other, which drives up the cost.
And it's about having to bear the clean-up cost when some customer who refuses to upgrade causes the shared hosting platform to become riddled with malware because this is inevitably happens, and drives up the cost too, and by a good deal too, because it needs to be handled by people with clue and people with clue cost money to hire.
So in case you're tempted to stick to your guns and moan that your fiver-a-month sharing hosting platform isn't offering you the long term stability that you want in order to continue to run your ancient and unsupported software, please bear in mind that there is a substantial cost to doing so, and that one way or another, someone needs to bear that cost, and that someone will be the customer because that's how supply of services works.
The only issue of relevance is deciding how that support cost is borne: whether through direct upfront cost increases, or by upgrading the platform, or by letting the platform rot and having customer attrition through neglect. Of these, the least bad long term option for the hosting companies is upgrading the platforms.
Otherwise I quite agree that it would be nice to have your cake and eat it.
it needs to be handled by people with clue and people with clue cost money to hire.
.... Doesn't sound very clueful for anyone to splurge lots of customer installations onto the same server without any isolation between them so that malware can spread, does it?. There is *always* one bad apple or even two on any shared service.
People who need old software hosted should just use VM's. Find a hosting company who supports full virtual machines, not containers, and the "malignant upgrade problem" will be gone.
When the tool chain used by a VM instance results in that instance being hijacked by malware, the hosting company has a severe problem on their hands. Having such instances can (will) result in increased resource demands that cost the hosting company real money even in the presence of resource limits. While I've never seen it occur (yet), their may exist legal exposure for the hosting firm due their negligence in failing to update their software and/or the activities of that infected instance. I rather suspect, along with more than a few in the security community, that's next. Right along with security regulations and strict liability for IoT as well. This ain't gonna be pleasant.
I have seen viewmodels with over 10,000 lines of code*. Apparently Single Responsibility Principle is the Single Responsibility Suggestion in some parts of the world. Windows as an example is circa 50,000,000.
* Yes, I wish I was joking too. And it very much was not defect free.
"If it ain't broke, don't fix it."
Indeed, rather than the issue of the OS being upgraded I think the bigger issue is of the software running on there never being backward compatible for very long!
I've seen it so many times when upgrading Linux systems, ou suddenly find that several functions / commands no longer work and break what you're using. It's like rather than fix the underlying fault, the dev's just decide to bin various functions and create new ones. Bugs needs fixing of course, but if they were treated more like black boxes there wouldn't be this issue. Eg, leave the function name alone, maintain the same input and output values (or add to them while allowing the original to still work) and fix the errors within the function itself. People using the function don't need to know what's changed inside the function, just that it still works. Most things in Windows work happily like that, for instance IIS, SQL etc, the commands you call work the same as before, and you're unaware that under the hood things have been rewritten and changed without breaking your code.
"If it ain't broke, don't fix it."
The problem comes when the rest of the world moves on, for example by retiring security cyphers that were in wide use at the time you implemented your software.
I don't like car analogies, but the demise of unleaded fuel meant you needed to apply some maintenance to older cars so that they'd continue to work in a post-leaded fuel world.
Correct me if I'm mistaken, but I believe you mean when "regular" leaded fuel was phased out because of pollution concerns, forcing some cars to get some serious work done to deal with the different fuel.
Question: Were some models forced into retirement because they couldn't handle unleaded fuel? This was during my childhood so it was all fuzzy to me, though I do know all the cars my family had required unleaded fuel. I'm just wondering if there were cars that required leaded fuel.
Yes, some models relied on lead/additive combustion products to lubricate cylender head valve seals/seats. If damage was likely for these types of engine, the components could be replaced. Alternatively, different unleaded additives could be purchased to mix with standard unleaded fuel to protect the engine.
I think that any mission critical software that carries a real risk to human safety should never legally be allowed to be done as long as the original vendor continues to trade. Support for Windows XP finished ages ago but MS recently released some critical security patches for it knowing that some crazy folk use it in hospitals.
"I think that any mission critical software that carries a real risk to human safety should never legally be allowed to be done as long as the original vendor continues to trade."
But what if the target is hardware, especially "set-and-forget" hardware that becomes unreachable once deployed, meaning you basically only get one shot at it?
"I think that any mission critical software that carries a real risk to human safety should never legally be allowed to be done as long as the original vendor continues to trade."
A fine concept, except within a decade or so all software will be provided by the likes of Capita, IBM, HP(E) and so on... is that what you really want to happen?
The sculptor Kathleen Scott had a definition for knowing when a figurative sculpture was finished. You walk round it looking for things to change - "Hmm - what if? - no". When you cannot find anything you think needs changing - then it is finished.
That doesn't stop you making your next one with some gained insights.
A bit like any project: version one does the job with some imperfections; version 2 is over-engineered; version 3 works.
"The sculptor Kathleen Scott had a definition for knowing when a figurative sculpture was finished. You walk round it looking for things to change - "Hmm - what if? - no". When you cannot find anything you think needs changing - then it is finished."
Problem is, the human mind is fickle. That's why there's the infamous interjection, "I change my mind." You may not find something you think needs changing, but sleep on it and you may find one...and another...and another.
I think you're thinking way, way too small. You must have watched the thousands of changed files go through on an O/S upgrade?
OK, you'd better contact the 50,000 developers around the globe working on thousands libraries and "sub tools" that the next level, the next level, and so on tools all rely on. Sure the venerable "grep" does not need any fixes but it relies libraries and sub-changes that affect it, so it needs to be recompiled. Now multiply that hundreds and thousands of times over the O/S.
Critical fixes are needed all the time, especially for Nasty-Net(tm) facing servers, it only takes one piece of scum to find out that you you have library revision 1.4.2.5.6 as opposed to 1.4.2.5.5 and it's vulnerable and you're screwed! Get to a point when enough fixes have to be constantly applied and it's easier to draw a line and set that as the starting point for the next version. Not every patch or upgrade it all about new toys, sometimes it's about saving your bacon from the bad guys, surely as a developers you cannot expect that everything is perfect as it is and software will be finished at some point?
As a systems admin, what gets me the most is that developers moan all the time that they need the latest features in order to code XYZ far more efficiently, and the second we say that it'll need a patch cluster or worse a full upgrade to get it, then we get a load of "earhole" about not touching the system. OK, fine but you can't have your cake and eat it.
Not quite!
I can't install the 1978 version of TeX on Windows 10 or even the WinXP/7 port and expect it to run!
This is the real problem, whilst the source code hasn't changed, what has changed is the OS platform, so TeX is only usable on Windows 10 because someone took the source and ported it to Windows and then updated the port and installer for each new edition of Windows.
Sympathise with this.
Today I used this little program, unmodified, and it still saves hours:-
// Withdrawable Plugs Program
// RG Started 29/10/96 last updated:- 4/11/96
// C++ - filename = wd.cpp
// IDE Notes :-
// set editor tab size to 3
// load project wd.prj
I am sure other readers will have more ancient stuff they still use?
Too many programs suffer from Constant GUI Overhaul Syndrome.
The recent trend to use "telemetry"[1] makes this worse, since if you find some functionality is under-used, you can conclude that:
1. Nobody wants it so it should be made less prominent,
2. or nobody can find it so it should be made more prominent.
In other words, there's never anymore an excuse to *not* tinker with the UI! O brave new world, that has such people in ’t!
Notes:
[1] Also known as "how I failed my statistician exams by using the Worst Biased Dataset Ever".
"1. Nobody wants it so it should be made less prominent,
2. or nobody can find it so it should be made more prominent."
Or 3, the feature is only used by the sort of power-users who are in the habit of disabling telemetry, and they must be punished for their disdain of telemetry by the removal of the feature. Mozilla are keen on this one.
When my SW was etched into Silicon, it was Very Much Done 3 months before it popped out the foundry. There was no hope of "fixing" anything without
a) Scrapping a million or more devices
b) Making new expensive mask sets
c) Waiting another 3 months
But when the cost is "download new version form the web", software quickly becomes evolutionary.
It's all down to your Definition of Done. i.e.. do you have one?
I think MOST software stopped being 'evolutionary' a while back, and became "we're millenials, it's OUR turn now to do it OUR way, screw you guys, we're doing it, stop being old farts and CHANGE!"
That's more de-evolutionary by my definition.
/me queues up some Devo music - "Go Monkey Go!"
Here here!
It is of course purely a commercial exercise to keep trotting out ever increasing bloatware but software companies forget that more features does not always equal better! Code bases get larger, more potential bugs and lets not forget; rendering completely usable hardware obsolete.
As mentioned constant UI revisions are simply bizarre:
Take the 3d effect UI and lets make the UI look flat
Nah don't like that lets make it 3d effect
Lets make top level menu items all Uppercase
No don't like that lets make it lowercase
and repeat Ad infinitum
I dunno, I reckon that upping the row limit in excel 2007 almost makes having to deal with the ribbon worth it. If it was still at 65k rows, the cry of being unable to open something in Excel would be hella lot more common in the office.
There's also some power user gunk and keyboard shortcuts which I think is more recent and very useful. Don't have a copy of '03 lying about to check though.
ping, or cat?
I'm sure there are a great many unix utilities that are 'done'
They were designed to fulfil a function, they do that, and so they are 'done'.
Of course someone will work out how to engineer a pong response that will overflow a buffer and feed you grandmother to the cat. But in real terms.... 'ping' is done.
The last git commit against GNU cat was on 17 June 2017 (so three days ago).
No behavior was changed but the code was changed to make use of the C99 declare-your-variable-in-the-for-loop construct. So apparently not yet done.
You may also be interested to learn that GNU cat has 17 command-line options (although many are GNU-style long-option duplicates). In all my years of UNIX/Linux use, I never recall passing an option to cat.
"No behavior was changed but the code was changed to make use of the C99 declare-your-variable-in-the-for-loop construct. So apparently not yet done."
hopefully not an example of what Linus might call "compiler masturbation"
https://www.theregister.co.uk/2015/11/01/linus_torvalds_fires_off_angry_compilermasturbation_rant/
I bought a shower, five years later it was so jammed up with limescale that I had to buy a new one and pay someone to install it, it would have been easier just to replace some of the parts (migrate to a newer, evolved version) but showers don't work that way so the old one became scrap and we had to do most of the work all over again.
People buy software and sometimes expect it to last forever on platforms that last forever, I don't understand why as it doesn't really work that way with anything else - food, consumer electronics, white goods, cars, even buildings need regular maintenance and updating or they become difficult to manage.
Thirty years ago I added a low voltage pump to the bathroom shower - it fitted neatly under the bath. About 10 years ago it failed - apparently the motor. Pump no longer made and no spare parts available. Bought the only one that seemed to have the same functionality. It still hasn't been fitted because it turned out to be too big to fit under the bath. Fortunately the downstairs bathroom's shower uses gravity. Now there is something that doesn't wear out over (my life) time.
a) We now have lots of beginners writing code for GNU/Linux who would otherwise have written shareware software for Windows which nobody would ever have wanted. The userspace is virtually overrun by those people. See systemd, much of FreeDesktop, and PulseAudio. There's even a term for it, Lennartware.
b) Some software projects have grown in size so much that they need companies to support them. One way to get an income in software is to sell maintainance contracts. That gives you a strong motivation to build your software in a way so you need to contract them. One example for that is Asterisk. You can run it without a contract, but the documentation is less than what you're used to. Most of it is some independent Wiki.
As I recall, Mozilla Thunderbird was proclaimed "done". It happened July 6, 2012.
The last sentence of the first paragraph here really does suggest that there's little room for improvement:
https://wiki.mozilla.org/Thunderbird/Proposal:_New_Release_and_Governance_Model
"As I recall, Mozilla Thunderbird was proclaimed "done". It happened July 6, 2012."
Despite being "done", it ended up with the "new, ugly" buttons that Mozilla uses for Firefox, including that *DAMNABLE* 'hamburger' menu pile of sewage.
Unfortunately, there aren't quite enough plugins for T-bird to get rid of the hamburger [though I was able to do it with firefox). I got the button appearance to change, but the DAMN HAMBURGER is STILL THERE! Grrrr....
[yet another example of "It SHOULD have been DONE, but then some millenial 'wizards' UNNECESSARILY SCREWED WITH IT and turned into THEIR way, and expected THE REST OF US to JUST LIKE it and CONFORM/COMPLY/SUBMIT"]
hmmm... that sounds like a good example of what the 'Millenial generation' of developers seems to be like: Instead of MIcro-shaft's "Embrace, Extend, Extinguish" they're wanting the REST of us to "Conform, Comply, SUBMIT!"
If you want to be able to run some "done" software for years and years then go with BSD. A clear separation between the system and user programs means you're much less dependent upon the whims of the vendor.
In reality anything that faces the internet is never done because not all of the bugs have been found yet. In the case of your Django site then you're looking at a heap of dependencies from glibc to openssl. Important to inform your customer early on that there will be a need for updates and upgrades (including the OS) over time and they should budget for this. You're doing no one a favour if you don't do this.
Not so long ago (well it seems like that to me) decades before the days of IoT your code was burn into PROM that was soldered into the circuit board in embedded systems with no way to ever change it. You needed to be quite confident your code was good before signing it off and agreeing the hefty PROM charge should be paid.
Once upon a time (around the year 2000), I started writing little utilities and tools, for my own use, using Python. This would probably be around Python 1.5. Most of this code survived without any changes through Python2. I thought most of this code was "done".
*
Then, sometime around 2014, Red Hat announced that in the future, Fedora would move to Python3, and that Python2, although still supported, would be phased out. On and off over the next 12 months I converted most of my small collection of utilities and tools...and no, I didn't do any improvements or tweeks....it was hard enough just to get the conversions completed.
*
"Done" is about the functionality you are trying to create. It takes no account of changes made by others, Guido van Rossum in this case.
While YOU may classify a chunk of software done, it may depend upon other software that is evolving and isn't "done". This automatically makes YOUR software NOT done as well. Sorry, this is a simple fact of life.
One can only hope that what you depend upon makes compatible changes which have little impact on your "done" software.
I suppose the only thing that is "done" is something that can't be altered in any meaningful way. The micro that controls my microwave oven comes to mind. Now we have "connected appliances" and again these are never "done", and we will all pay the price.
No, I don't need a connected refrigerator with a silly display. I need a box that keeps my soda & beer anti-warm.
Reality set in. The premise that sometimes software is just "done" makes sense, But in the real non-computing world I have come up against plenty of examples of needless changes. Each time we need to replace a bit of crockery they've changed the design, just enough to make the new item look different. Or we need to retouch the paint and they've replaced the colour with a new shade that is just a fraction different (over and above batch variations). Or a door catch barrel that is a slightly different shape, so that the cavity has to be made slightly bigger in places. And so it goes.
"Or we need to retouch the paint and they've replaced the colour with a new shade that is just a fraction different (over and above batch variations)."
I thought most paints these days were tinted to order. Just bring a scrap of the color you want and they can mix you a quart of gallon to match it.
Sadly paint changes colour with exposure to sunlight so even if you had some of the original paint to use it wouldn't look quite the same. And if you mix some new stuff up to exactly match the chances are it won't quite match in a month. Not that I notice, but my wife will point it out.
The installation of the shower is done, just like a software installation can be done.
What comes afterwards is maintenance, if the shower breaks you have it repaired, if the software breaks you have it repaired in the form of patching.
Sometimes the shower may be deemed beyond repair (parts are no longer being made, cost of new parts exceeds cost of replacement etc) and so the shower might be replaced with a different unit instead of repaired, which is analogous to upgrading to a new version.
Windows XP didn't stop being secure, it never was secure we've just never been aware of all the security problems it has. There will always be 0day exploits out there for which even currently maintained software is not patched (see how long equation group had their exploits before they got publicly disclosed). Just like any other device, the manufacturer is no longer willing to carry out repairs or produce new parts.
The analogy with a hosting provider is slightly different, it's more like you didn't purchase the shower, you're renting it (eg you rent an apartment which includes a shower)... It's no different than if your landlord decided to replace the shower in your apartment. If you rent, you've less control than if you bought.
Also when you install a given piece of software you should be aware of its lifecycle up front, you know the vendor will provide security updates for a specific time and you know what your options are after this time. All of these things should be taken into account when procuring software, and you should demand the same level of predictability when procuring a service too.
You supplier no longer wants to support the Centos version and toolset your legacy software runs on.
No problem. Just find another supplier and move to them. There isn't one? First tap of the clue stick.
No problem. Find a bare metal hosting company and set up and maintain your own platform with the OS and toolset you need/want. Too hard/expensive/time consuming? Second (harder) tap of the clue stick.
What is the cheapest and most practical option? Upgrade the software to run on a current platform? Oh, you just did that?
Then stop fucking moaning! The turtle moves!
There are quite a few situations where software is declared "done". Anything on a satellite or similar spacecraft will be declared finished some time before it is put on top of a rocket and sent to space. Nobody is going to be allowed to update anything remotely, in case something is broken in the update.
Similar examples are to be found where lives are at stake. Any Safety of Life system may not be upgraded or otherwise tinkered with, other than through a highy controlled process - typically, a new procurement. Systems involved in the production of pharmaceuticals are similar - once a company has spent several years and several million getting approvals from regulatiry bodies, they are going to let someone tinker with anything.
The trouble with using analogies is there is always some anal literalist who likes to home in on a difference of detail between the situation and its analogy (there has to be one, or it would be an identical copy, not an analogy) and uses the necessary difference to argue that the analogy doesn't apply. No matter that the difference has no relevance, it's there, and so it's a perfect smokescreen for setting up a pedantic distraction.
I loved this article; it is so true of the software industry. On a much more trivial level, I get irritated with Microsoft "improvements"/"updates" to its Office Suite products. The worst one is the XY chart in Excel. The old versions of Excel were brilliant. You just pressed F11 on your selection and up popped a perfect chart. That still is the best method in my view. They had a brilliant Chart Wizard for doing all the customising and topping and tailing. Done away with. Around 2007 a ghastly version of charting in Excel came out. You had to click on one of six or eight panels on the left to choose your chart type, and none had a label in words. The diagrams didn't make it at all clear what was going on. If you chose the top left one, you actually got an incorrect chart which defaulted to a cumulative plot instead of a straight one. I was doing the IT skills part of an accountancy course at the time and when I pointed out to the class tutor that no-one in the class had produced a correct chart, including the tutor herself (there were clear contradictions between the tabulated data and its chart display) she nearly had a heart attack. I seriously wondered if the error might take in the examiners as well, so that if anyone produced a correct chart (by F11) they would be marked wrong.
Worst of all, the New Improved Excel doesn't do proper XY plots any more. The old Excel pre 2007) was brilliant at plotting one data set along the X axis and the other on the Y axis - great for illustrating scattergrams of partly correlated data. But now, if you choose the so-called XY option, what you get is a double graph of both sets of data on the Y axis against incremental steps along the X axis. Useless, and so infuriating. Public and work computers have all got the "up"graded Excel loaded, and I can't do proper XYs any more. Is there any way of loading the old Excel and getting our XY plots back?
io_uring is getting more capable, and PREEMPT_RT is going mainstream The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
