ARPANET pioneer Jack Haverty says the internet was never finished Early-internet pioneer Jack Haverty has described the early structure of the internet as experimental – and said not much has changed since. Haverty was a protégé of Professor JCR Licklider in the early '70s, when he worked on the then brand new ARPANET. He is widely credited with developing File Transfer Protocol, the RFC …
... The Internet is nothing more than an insecure research platform gone mad.
When we built this thing, we knew there were issues, even pre-BARRNet and NSFnet. It was not ready for prime-time back then, and it's still not ready for prime-time. We tried to keep the corporate world out of it because of this, but we we naive. We failed. We apologize.
I'm 99.9% certain that I can speak for us, especially seeing as TINU.
Yes, and TCP may not be quite finished, but it still works pretty darn well at Gbit rates despite the imperfections. As jake knows well, assuming it's the jake I think it is, principles were set in the late 1970s that have worked brilliantly. Some of them weren't even written down until 20 years later, if at all.
This post has been deleted by its author
We've got the results all over the hills here in Sonoma. Their mums are scarfing the mustard that grows between the vines like there is no tomorrow. 75F/24C, light breeze, the volunteer tomatoes and peppers are already showing their first true leaves ... You'd almost think it was Spring :-)
Yes. The mustard will start seasoning the lamb from the inside. The newborns only nibble on it a bit (imitating Mom), but as they get older they will eat more and more. By Easter (harvest) the taste difference between barn-raised bottle lambs and field-raised lambs will be quite noticeable. (Also, the ones with Mom out in the fields will be meatier, the ones hand-raised in a stall with a large paddock will be fattier.)
Around here, Saltmarsh Lamb (and Mutton) is the thing. While sheep in the UK are generally outdoor reared - only coming into the shed to lamb, and for some breeds not even then - Saltmarsh grazing with (from the website)
a natural abundance of Samphire, Sorrel, Sea lavender and Thriftreally does affect the flavour.
And the cost.
M.
I go one better ... remember the vines that the mustard grows between? I sell the complete package: one whole lamb, packaged for the freezer (or pre-frozen, if I have to ship it), and a case of Merlot from the grapes that grow in the same field. The Yuppies buy me out before the lambs are even conceived ...
> There is no such thing as a prototype. There is only the final production version! This may be replaced in the future
There will be an attempted rewrite for v2 which will be abandoned due to scope creep. v2 will be skipped and the prototype will go back to being v3.
"Sorry for the downvote"
So change it back. I just changed yours from up to down to up to down and finally left it at up. I noticed one could change one's opinion years ago ... the only problem is that you can't remove a vote completely if your cat decides to make up for its lack of thumbs ...
This will, I suspect, amuse and horrify in equal measure, but I recently prototyped some abstract modelling code in fortran, because (a) I'd been fortranning on some moderately creaky code immediately prior, and (b) I wanted to update myself to all fortran's newer module features.
Then once I'd got a handle on the best architecture for whatever the research problem actually turned out to be, I was to re-code it properly in python, with classes and all that. I thought that a rapid and disposable fortran prototype would be more useful and less annoying than iterating through several clean restarts of python.
Although I did indeed start the python version, it then turned out it would have been way too slow, so now the fortran version is the main code. But at least I'd got the fortran code's design good-enough so as to not need to restart it from scratch! :-)
Given how much computational code is written in Python but essentially running in Fortran, the usual approach is to prototype in Python but write real thing in Fortran or C++, or use any of the existing libraries.
Python and other similar languages have proved a godsend for all kinds of work because they let talented scientists do stuff they'd never otherwise be able to do without first getting a Comp Sci degree first. And at least 50% of this is down to providing bindings for decades of excellent Fortran and C++ code…
Whatever the current state of any tech there is never going to be a demand for something better. Until the marketing suits present it to us. For the simple reason that the customers don't know what the potential, the justified expectation, is.
Tech, for almost everyone, is simply magic with bells and whistles. So as long as it provides the advertised "user experience" no one is going to be satyng "Why is it like this?"
No one, that is except El Reg commentards. And let's be honest, no one listens to us.
One of the enduring things about the internet is that it has evolved within the original unfinished constraints.
Hugely different access methods from dial up, serial, leased lines, copper, fibre, microwave, WiFi, 3-5G, satellite return (uplink via a cable) to now bi-directional satellite. Who knii oh we C what access methods will be available in the future.
The way we use the protocols have also matured. People have reinterpreted the original protocols and we now see huge performance improvements, security, availability and that has fuelled its growth.
I remember being laughed at when I said phone companies will use internet circuits to complete calls instead of the pstn, then we had BT’s 21cn which did exactly that but to local exchanges.
Skype ushered in a whole new internet use case that is now so ubiquitous people don’t understand how it happens they just care it does.
If a finished internet was delivered, all the mega corps would have nobbled it to protect their businesses. There would be no voip, no on demand video, no music, just no media in general. In fact fastest speeds would be a fraction of todays but perhaps those that paid more to transport their content would get faster speeds than other services.
"One of the enduring things about the internet is that it has evolved within the original unfinished constraints."
I'm not all that certain you understand what the network we call "The Internet" really is.
"from dial up, serial, leased lines, copper, fibre, microwave, WiFi, 3-5G, satellite return (uplink via a cable) to now bi-directional satellite."
The network doesn't see any of that. It's all just wire, and wire's wire (even when it's pigeons). We built the thing to be wire agnostic. In fact, it was supposed to be hardware agnostic, with all of TCP/IP handled in software.
"There would be no voip, no on demand video, no music, just no media in general."
That's all above the level of the network. TCP/IP quite simply doesn't care.
"In fact fastest speeds would be a fraction of todays"
The network doesn't care about speed. Humans do.
I'll have to agree with that. The initial design is what allowed all those technologies to be used. Note that word: design.
Just because the design wasn't complete doesn't mean it was garbage; it was meant to be built on, and build on it we have. It has just been far from elegant, and the designers probably had more elegant solutions in mind after they'd gained some experience with what they'd built, but it became general-use before they were ready for that to happen. I can certainly see evidence of the failure to evolve the internet in our failure to even enable widespread adoption of IPv6 after all these years of it being ready for production.
I don't believe IPv6 is even an option with my current ISP unless you are a full-on business customer with fully synchronous bandwidth and your own Cisco hardware on premises to handle the fiber link.
“ I'm not all that certain you understand what the network we call "The Internet" really is. ”
I'm not all that certain you understand what I wrote.
I mentioned the various access methods to highlight that it works across all of those plus new ones that haven’t been invented yet.
“Who knii oh we C” Should read “and who knows”
The complete sentence to make it clear to you.
— Hugely different access methods from dial up, serial, leased lines, copper, fibre, microwave, WiFi, 3-5G, satellite return (uplink via a cable) to now bi-directional satellite and who knows what access methods will be available in the future.
“ That's all above the level of the network. TCP/IP quite simply doesn't care.”
It doesn’t care because it was never considered in the design, design the internet today with input from the global technology companies and they will insist on putting things in that will have a bearing on their niche.
If you used all my words in context it is clear and obvious what I’m talking about rather than your picking bits out of context having a moan and trying to belittle me.
To be clear here is the full paragraph your moaning about but actually agree with.
“If a finished internet was delivered, all the mega corps would have nobbled it to protect their businesses. There would be no voip, no on demand video, no music, just no media in general. In fact fastest speeds would be a fraction of todays but perhaps those that paid more to transport their content would get faster speeds than other services.“
Good chat Jake
No, it doesn't care because that's not "the internet", that's just your local device and network access method, your local network (even if that is actually an ISP you connect to, in whatever way) then connects to "the internet" (backbone network) and from that you can access anything else that is publicly accesible from "the internet".
To put it another way; there is only one way to "access the internet" and that is to implement a TCP/IP stack on a device that can communicate with "the internet". All those "access methods" you're talking about are ways for your end devices to connect to other devices that, can themselves ultimately connect to the backbone devices. They don't connect you to "the internet" per se, they connect to another device, which may or may not be networked itself, and so on, and so on. "The internet" was not designed as some monolith that you connected to in a certain was, indeed as 'jake' has implied several times, it's debatable it was really designed, so much as thrown together and played with, then left standing to play with some more. It's fundamentally a collection networks, with their own topologies, devices, standards and so on, that also connect to other networks - which they do with TCP/IP.
This used to be much more visible when local networking used protocols that were not TCP/IP and only used TCP/IP to reach "the internet". Those days are not all that far away for PC users, where IPX/SPX, for example, was quite common.
Tsk, man, the kids that think the internet began in the '90s or later don't remember IPX/SPX (Netware, I believe), or token ring, or point-to-point dial-up, or BBS's, or any of the early technology that led to the need for TCP/IP to pull them all together. In fact, I suspect many of them don't think "the internet" existed until there were _games_ available that used it. *LOL*
...people don’t understand how it happens they just care it does.
This phrase can accurately be applied to almost every user of technology throughout time. I say "almost" because some of those who create and support a given tech also use it and they might understand it as well. The rest have no clue nor care.
I think there are a lot of levels of understanding.
Many will understand how to implement firewall policies in checkpoint or PA but 99.999% will have no clue as to what’s happening under the gui. Neither should they care as that’s why they buy the product with maintenance. They are not expected to dive on the cli and do stuff (should be more if a rarity on checkpoint these days but a common thing in older versions).
Max Verstapen is their reigning world drivers champion*, he’s an expert at driving last years RedBull, he has no clue about the details of the aero or engine or brakes etc etc etc.
He truly just cares those components work.
Not intending to start a flame war but we have had protocols, RFC's and STD's come and go over the last 40 years (SMTP RFC came out in 1982... damn I'm old), yet we are still complaining about ipv4 address depletion for over 20 years, we are still using it, and will admit even though I'm seeing more ipv6 addresses in use I still get an uneasy feeling when I see one vs the good old dotted quad...
Am I passed it and holding onto the ideas of "it was better before" or should we start thinking about a better internet protocol?
What would be a better internet protocol.
Ipv6 has changed a lot yet it still hasn’t really gained traction.
How about Google’s version or M$/Meta/insta or Apples or Twitters or Amazon or the Chinese or Russian alternatives. All powerful business will want to optimise things for their use cases.
I like ipv4 because people have had to engineer around it’s limitations and come up with some howlers plus lots of very useful solutions.
We all know ipv4 is limited but instead of just increasing the address space ipv6 mandated specific ipv4 use cases be eliminated, like NAT.
The truck they missed is that people would have moved to ipv6 then transitioned off of some legacy concepts.
There is no point building a new version then forcefully removing the comfort blanket while trying to persuade people to use your new shiny shiny.
Wtong. It was said that NAT would no longer be NEEDED, which is not the same thing.
NAT for Ipv6 exists. Both via the analogous NAT66, and the better NPTv6
(There is also NAT64 for NAT related ipv4 <-> ipv6 translation)
If they had just "increased the address space", people would be howling about how there's this whole new protocol to implement, and it doesn't address any of the IPv4 inefficiencies.
Just about everything in IPv4 can be done in IPv6 (even when better methods exist)
The only thing I can think of that doesn't exist in IPv6 is arp/rarp - IPv6 uses local multicast rather than local broadcast, but the result is the same.
Nothing in my post but facts, yet I get 4 downvotes and no replies.
It's getting more like the Daily Mail / Mumsnet here everyday.
What's wrong? The facts hurt your feelings?
And before anyone says it; if I cared about downvotes, I wouldn't be posting this response.. It's just sad to see such blinkered and irrational reactions on here.
You are probably being downvoted because of your vehement use of "wrong", implying that everything posted was incorrect.
With that said, many of the IPv6 advocates have, in fact, called for the (eventual) elimination of IPv4 use cases, like NAT.
Frankly, I like NAT. It's a really, really useful tool to have in the ol' box, for lots of reasons. Anyone who calls for its removal either isn't an actual admin, or has completely lost their mind.
so nat66 is stateful and the closest to ipv4 nat yet there is no rfc for nat66.
npt66 effectively retains the upper most significant portion of the address and swaps the lower least significant bits and is stateless.
in this day and age, for security we like to break the path, proxies and Load Balancers are useful for this, this can be done just as easily in ipv6 with the same but the point is that direct end to end comms is no longer desired.
Nat is extremely simple, not complicated and useful in addition to firewalls and achieves in concept what going through a LB or proxy does without the complexity.
IPV6 actively discourages NAT
see RFC 4864 which nicely details how IPv6 is designed to not use NAT
https://datatracker.ietf.org/doc/html/rfc4864
The reason IPv4 still dominates is largely down to IP address depletion not affecting everyone equally. If IPv4 addresses had been allotted equally be country initially, the US would also have started to run out. IPv6 solves some of IPv4's problems, including address range, and is seeing increasing adoption within the networks and service providers with end users hardly noticing: their mobile network is probably already IPv6 and more and more providers are now 6to4 rather than allowing users to go 4to6. The transition will continue to be gradual and may never be complete. But this is true in other areas: altitude for planes is still in feet, speed of ships is still in knots, and some engineering is still done in thous…
> Am I passed it and holding onto the ideas of "it was better before" or should we start thinking about a better internet protocol?
It wouldn't be the worst thing to just add another number and a dot on the start of it and go from there, with quads being interpreted as being prefixed with 0.
Has many downsides, but easy to eyeball.
I have thought this myself. It would solve the problem, and would be easy.
Impact would not be felt for a few years until people with out of date networking gear start being unable to reach certain addresses.
Best of all I won't have to try to hold an IPv6 address in my head, which is clearly impossible.
It's not the IPv6 address size that bother me... I'm sure my brain cells will "rewire" to be able to read them given time, it's the fact that my eyesight is nowhere near as good as it used to be and the blur created by a chain of 3 semi-colons looks more or less the same as that from a chain of 2 or 4 semi-colons.
"I'm sure my brain cells will "rewire" to be able to read them given time"
I've been working with IPv6 for about as long as it's been around. My brain cells have not yet rewired themselves, which in and of itself is an anomaly. Usually this kind of thing only takes a week or so, a month at the outside. But it's been decades now ... do with that what you will.
"my eyesight is nowhere near as good as it used to be"
Indeed. This might be a part of the reason for the above ... but I suspect that it's more because ordinary humans don't have that kind of internal interconnections. The synapses just aren't wired that way.
I realised that (colons) just after my edit window expired. Hoist by (my own, but not) pedantardary!
I still find it incredibly difficult to read IPv6 addresses. One or two colons still blur a bit, and the compression rules allow for one address to be compressed to several equally valid compressed values.
I also find them quite hard to visualise as bit patterns. Again, that may be just brain rewiring that's yet to happen.
Jeeezuz, how many times does this come up.
>Impact would not be felt for a few years until people with out of date networking
>gear start being unable to reach certain addresses.
If you add any extra fields to the IPv4 structure for addresses it doesn't magically work.
Everyone's equipment has to speak your IPv4 "yet another shitty hack" edition or it doesn't work.
>Best of all I won't have to try to hold an IPv6 address in my head, which is clearly impossible.
Do you realise this is like saying we shouldn't have 64bit address space computers because remembering 64 bit addresses is more difficult for the tiny amount of people that need to mess around with stuff at address level? Address are for machines, DNS is for humans.
Can a rule be added to the register ToS that says something like:
"If you have a better idea for addressing the IPv4 address space issue before you can write a smooth brain comment about it you have to push your modified IPv4 stack to github and show it actually works properly with all IPv4 hosts."
It wouldn't be the worst thing to just add another number and a dot on the start of it and go from there...
You're kidding, right?
Have you never seen anyone develop software before? The amount of existing code that makes assumptions about the length and format of an IP address is astronomical. Suddenly changing the format like that would be a recipe for disaster.
We're talking about the sort of quality code from developers who forced Microsoft to skip Windows 9 because they were worried software might assume it was running on Windows 95/98.
Yes, I get that's what you're saying.
What I'm saying is if IPv4.5 addresses look like IPv4 addresses a lot of software will break because of the bad assumptions it makes.
By making IPv6 addresses look completely different, poor parsing algorithms will more likely fail immediately instead of partially working and then failing for some other reason. (Like not being able to connect to a non-existent server because the address was only partially parsed.)
>Have you never seen anyone develop software before?
>The amount of existing code that makes assumptions about the length
>and format of an IP address is astronomical.
IPv4 stacks aren't based on assumptions. The length of an IPv4 address is part of the specification. And the specification says it 32bits. You can't just add bits to it because the specification says it's 32 bits, not a string of characters or a multiple of 8 bits with conituation bits or anything like that. Just so everyone is clear: An IPv4 address is not a string like x.x.x.x. It is a 32 bit number.
And if you think it's just software that considers an IPv4 address as a 32 bit number you are not seeing the whole picture. There are hardware packet processing engines that also operate on the IPv4 header and expect a 32 bit number for the source and destination addresses and will ignore whatever you try to do that add more bits there.
Frankly, I haven't seen an actual, RealWorld reason to change from IPv4 to IPv6 ... Yes, I know, we're running out of addresses. But we're not out yet, despite the hand-wringers moaning about it for the last 35 years or so. IPv4 addresses change hands every day.
IPv4 will be with us for the foreseeable future ... and I'll probably stick with it for my personal systems about as long.
With that said, most of my customer facing kit is IPv6 ... Contracts demand it, EVEN IF their existing kit is perfectly happy with IPv4 ... and in fact, I'm asked to "upgrade" IPv4 kit to IPv6 kit on a fairly regular basis, even when I can demonstrate that there is no need to do so at the moment. Whatever, it's your money.
There are some added features, but in reality IPv6 was a very conservative upgrade from IPv4; more radical ideas were ditched because of conservatism among the equipment and software manufacturers in the mid-1990s.
IPv6 utilisation is around 36% of the total now, according to Google.
But yes, the IPv4 standard is still called a Request for Comments.
There are certainly elements of this, which is probably why uptake was initially so slow. However, nowadays you could consider adoption an example of Pascal's wager. At some point IPv6 may well be dominant so why not get on board now? If you're going through a network upgrade anyway, being able to run IPv6 outside your network isn't going to hurt.
Pascal's Wager was flawed in conception. He didn't take into account the concept of multiple gods, nor that his choice of god might be the wrong one.
Which divinity should one pray to, pray tell? Mankind has had tens of thousands of them over the years, your chance of picking the right one out of the bunch is slim to nil. So go ahead and pray to your god ... but you'll feel a right silly bugger if there IS an afterlife and Ra wants to know why you've been praying to a false idol. Or worse, Kali ... or Wōden. Surely the proverbial "thinking man" should say "I don't know", and then apologize when arriving on the other side (assuming, of course).
For you xtians out there who claim "one true god", which one of the two in the bible is the one? I rather suspect that the god of the OT would be really, really pissed if you are worshiping the God of the NT, if they are, in fact, different entities (and they sure have different personalities!).
The article only hits at what features the internet is still missing in the eyes of its creators.
As far as I can see, the internet is one of man's greatest creations, with billions of people viewing pages, applications and audio and video on it on a daily basis. It's simply amazing that it will works as well as it does. It makes you wonder: what could still be improved?
Not the part where you want to get data from a different system or put your data onto one. That requires the machine-to-machine bit. That is one of the levels of the internet, but if you want to limit the definition only to OSI levels 3 and 4, it's still critical to lots of systems and has seen many refinements. It's not done, and it never will be done, but it's impressive how well it tends to work.
"The rest of his career more than makes up for it, though."
Arguably, without FTP, or something like it, The Internet (as used by consumers) would never have been built. So while I agree with you, it's probably not for the reason most might think :-)
"Arguably, without FTP, or something like it, The Internet (as used by consumers) would never have been built."
The key part being "something like it". I don't have much objection to FTP, which is probably because I never had to write an FTP client, but if he hadn't made that, someone else would have used a different protocol to move files. It's kind of obvious that, if you have files and you have a communication system that can send data, you will eventually want to put those together. FTP is important historically and in many modern-day situations as it's one of the oldest protocols commonly supported, but many of the protocols used by those consumers have superseded its function in a different and often simpler way.
First you go to school for three to eight years, majoring in network programming with a strong minor in operating systems, and then you'll not only know what is needed, you'll actually be equipped to do something about it.
But you won't, because like the rest of us, you'll realize that what we have, while flawed, not secure, and arguably not really fit for purpose, it is good enough for entertainment purposes :-)
I'll bet if you'd let those engineers keep it "in development" it would still be "Wait: it's not done yet. We have this long list of things we still have to figure out" today.
Robert Watson-Watt couldn't really fit radar on a WW-II fighter, so he justified his choice of a non-optimal frequency to make it small enough with "Give them the third-best to go on with; the second-best comes too late, and the best never comes"
"I'll bet if you'd let those engineers keep it "in development" it would still be "Wait: it's not done yet. We have this long list of things we still have to figure out" today."
But it IS still in development, and there IS a long list of things that need figuring out. But during the meanwhile, as bits are added. taken away, and/or modified, the whole keeps plodding along, quietly doing the job that it has evolved to do.
Internet developer does a professional Rip Van Winkle, comes back in a few decades and finds that on the one hand, no notable progress has been made to the underlying tech. On the other hand, everything still works pretty well.
A few parts horrifying and a few parts inspiring. The fact that things have held up as well as they have says much about the skills of the early developers.
Unfortunately for us, the next Jack Haverty is probably inventing the latest tracking tech for Amazon or Google.
"The fact that things have held up as well as they have says much about the skills of the early developers."
And the language it was written in (and continues to be re-written in) has weathered time rather well, too, despite all the kiddie detractors claiming it's useless, archaic, dangerous, and etc.
I would claim to be one of the very few people who have had to implement it from scratch. Most people just use it and don't realize what an utter crock it is, an astoundingly inefficient, unstable and mostly unnecessary communications protocol that because of its wide, and unthinking, adoption has survived only because the brute force power of the network hardware has disguised just how bad it is for most users.
It does have its uses but they're primarily emulating a serial terminal.
The biggest headache with the TCP/IP suite (at least in my opinion) is also one of its strengths ... that being that it is Open Source. Most of the protocols were developed by one or two (three, maybe four) Grad students hacking away until either it worked, or they got tired, or they graduated and moved on to other things ... at which point, another grad student (or two, or ... ) picked up the freely available ball and started playing with it.
As a result, the rather complex whole looks like it wasn't designed (because it wasn't!), rather that it grew, or evolved.
This makes it rather difficult for the neophyte to grasp as a whole in anything resembling a hurry. It's not cohesive, and in some ways is rather incoherent.
On the other hand, because it's both Open Source and built of little chunks, divide and conquer works. The very same neophyte can pick a small portion and study it, and perhaps even improve it.
A better alternative? If there were one, the pR0n industry would be using it. They are not, ergo one does not (yet) exist.
>It does have its uses
It adds data integrity to the link. That was basically the original requirement. Nothing more.
That sounds easy, but it's not. It's ludicrous to question its usefulness.
The messy stuff is options and optimisations, which have accumulated over the years. (And of course, the mess and flakiness added by people who don't know what they're doing- just poor libraries, i.e. bad coding.)
"At one of the quarterly meetings Vint Cerf came in and dropped a bombshell on us: he said TCP had become a standard."
It sounds like he went for 'scale' early, to beat anyone else to it: and that's a very modern mantra.
In all fairness, there have been projects I've worked on that are well north of a decade old, and there were still new requirements, new ideas, and new concepts being added to the project "todo" list.
Unless someone imposes a deadline, techies will always have more and better/cleaner ways of doing things, because code and systems evolve, they aren't just cast in stone from the start. Unfortunately, the internet has not been allowed to evolve even so much as to embrace IPv6!
"Oh sh*t, another beautiful day in California!"
Probably 10 degrees (F) cooler than yesterday, and a mix of high and low clouds. Was foggier than shit as we ran under the Golden Gate this sunrise. Couldn't even see the bridge. (Sausalito to Bodega Harbor was cool (50F) and damp, seas of 8-9 feet from the West 14-15 seconds apart, wind 10 MPH from the West, all the way. Nice morning for a putt.)
"When's it gonna rain?"
Should be around half an inch here tomorrow (Thursday). A foot of new snow at pass level in the Sierra, probably a foot and a half on Shasta and Lassen :-)
"At one of the quarterly meetings Vint Cerf came in and dropped a bombshell on us: he said TCP had become a standard. Our immediate reaction, or at least my reaction, was 'Wait: it's not done yet. We have this long list of things we still have to figure out'," Haverty recalled in his speech.
I'm reminded of talking to somebody who participated in the efforts to produce the first standard for C++ which emerged in 1998:
"They don't want it right. They don't want it complete. They want it Tuesday."
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
