Remember SoftRAM 95? Compression app claimed to double memory in Windows but actually did nothing at all One of the most consistently interesting and entertaining Microsoft blogs, Raymond Chen's Old New Thing, recently covered the dissection of a best-selling bit of software for Windows 95 – SoftRAM 95. There are few lessons about modern software in there as well. What SoftRAM 95 claimed to do was software memory compression – on …
I think my first example of "software that does what the OS should have anyway" was a piece of software that would resume downloads when someone rang your mum every 20 minutes and killed the internet connection, making that 3mb mp3 take days.
That thing was amazing considering we take resumed downloads for granted now
And before you say "Disable it", for me at least, I was only allowed to do 8+ hour online sessions if I kept call-waiting on, in case someone needed to contact my parents!
Unfortunately, I was stupid enough to mention this to friends who took great delight in kicking me offline whenever I was in IRC....
This post has been deleted by its author
I'm sorry I forgot to mention that in the US most of our parents had Call Waiting™ (so Aunt Kate could interrupt Mom's call with Aunt Terri) that would interrupt a call with a beeping tone, Not if the four tunes for a modem and cheerfully freaking your modern out.
*70 was how you killed that, and you added it to any dialing software worth a crud as a prefix.
In case it's useful to anyone who reads this, if you have ADSL filters and are still losing connection when someone phones you it's likely you have a high resistance in the crimps or lines somewhere. If after you replace the filters and narrow everything to just your wall box, a phone and the modem, it persists you have this problem.
The only fix for this is to get a line test done for high impedance and a phone tech to actually fix it. It should be free for most anyone if it really was a high impedance error.
I don't remember the sequences, but with BT there was a digit sequence to enable it, a sequence to disable it, and a prefix sequence you could put before a number to disable it on a per-call basis.
No adaptor needed, but for many people with only one phone line, the reason not to disable call-waiting wasn't a technical one!
Was that not when xmodem was replaced by ymodem/zmodem?
I recall some friends with BBS deliberately not offering that as protocol to users who only came to download (the idea was that you occasionally would contribute something too). By sticking to xmodem they could punish the mass downloaders by killing the link just before the last few bytes had gone down the line, thus invalidating the whole download..
(did you really think the BOFH was a recent idea? :) ).
He's right though. The registry doesn't need cleaning if you're doing a fresh install of Windows every few years. Otherwise, the amount of crap that accumulates there can be a sight to behold.
There's a reason why all these so-called registry cleaners are so widespread - there's a market for it, because issues with the registry can and do cause problems. This "feature" plagued Windows ever since its introduction, and while it's perhaps less horrible than it was at one time, it's still not anywhere near good by any stretch of the imagination.
I seem to recall the "Registry" wasn't even supposed to be a thing, one department at Redmond developed it for their internal use, and then everybody wanted to use it... and then the genie was out of the bottle. I can't seem to find the story anywhere?
Why is he right? You'll have to explain why an application would care that the registry has lots of entries in it. If the application cares about other parts of the registry than is relevant, that would seem to be an application problem, not a registry issue.
If I get a new appliance to replace an old one (ie, upgrade an application), why should the fact that my basement is full of crap be an issue at all to said application?
"Why is he right? You'll have to explain why an application would care that the registry has lots of entries in it."
Exactly right. A "messy" registry will make the system boot up a bit slower, as the larger hive takes additional time to load (inconsequential now, thanks to SSD's). After that, the excess data is simply not accessed and therefore has almost no impact, unless and until the extra entries point to a system module that is no longer needed to run but does so anyway. Blame the uninstallers in that case, incompetent programmers.
The problem has always been that. Application/installers write to the registry and they should remove their entries when no longer needed. The operating system may not know if the entries left are to be reused - i.e. reinstalling an application that can find and reuse the entries, or not. It's just you can find cruft left behind in /etc and other directory because the package remove code doesn't clean up everything, often leaving behind user created or modified files. Most developers are more interested in writing installers than uninstallers - after all you should never remove their incredibly useful and state of the art applications, shouldn't you?
Maybe the system could have kept track of added entries, and garbage collect them after some time or on user request. In some ways, Windows Installer does exactly that. Other setup tools - i.e. Inno Setup - require to setup the uninstall phase too - taking also into account keys that the application itself may create after setup.
A stable system which doesn't see software installed and removed often usually doesn't suffer from Registry issues - installing/removing a lot of badly written installers and software may leave cruft behind.
Yeah, because building systems the blindly trust 3rd parties have worked out so well for us. Let's just run everything as root because "why shouldn't every app developer be responsible enough to do that without shooting themselves in the foot and the user in the head".
There is a reason people are talking about zero trust and least privilege.
Sigh.
Running as root has little to do with reg bloat. The registry has two parts, the machine hive you need admin rights to write to and the User hive. Every user has one of these and writes to it in their own user context. Some reg bloat can happen in the machine hive (several Windows bugs have caused this in the past), but it is normally shitty apps writing to the user part.
If an application is bloating the user hive, then it doesn't matter if you run as root or not. Any sane business will not let users run with admin privileges, but will still need to keep any eye on the registry.
> It's just you can find cruft left behind in /etc and other directory because the package remove code doesn't clean up everything, often leaving behind user created or modified files.
Most applicatoins do this intentionally. If the user has modified a file, then the app doesn't know if the user needs it or not, so errs on the side of safety and leaves it behind.
Take, for example, samba. If you install samba, modify config files to enable a whole heap of network shares with complex permissions. If uninstalling samba deleted these files, it would mean you've now lost all your network sharing config that you were going to use with the new version that you've just installed.
No, an 'uninstall' should never (by default) remove user-modified files, as maybe the user modified them for a reason and still wants them afterwards.
No, an 'uninstall' should never (by default) remove user-modified files, as maybe the user modified them for a reason and still wants them afterwards.
Now I know this inspires horror in many, but a solution for this was actually discovered ages ago, known as actually asking the user. The whole problem starts with the assumption that a developer somehow magically knows the deployment environment better than the user (or sysadmin) in question.
If in doubt, leave the option open, even if you have to stick it behind an "advanced" or "here be dragons" barrier.
As I explained before, that what happens with the registry too. The OS can't know if what has been left behind was left intentionally, or not. So, registry cruft it's not because the OS can't keep the registry clean itself. If it's a configuration file or a configuration key, it's the same issue.
Good uninstaller should ask the user what to do with anything that might be left here. Thus blaming Windows for what is a responsibility of application developers is a tad unkind, while the same is deemed a very good behaviour on another platform....
And the same it's true for the cruft left in the user home and AppData folders - it looks many application can't really clean up the mess they did when uninstalled.
Amen to that.
Many years ago I was talking to one of the windows team who told me the original use of the registry was limited and it was in particular intended for information that the OS or multiple applications needed rather than any individual application's needs. Then people started sticking stuff in the windows registry that should have been in a .ini file or other application specific config registry and from there it all got totally out of hand.
But it's not just the registry. It's files and entries in files etc. Many ages ago we tried a little exercise. We took a snapshot of the registry then installed an application then uninstalled the application and compares the snapshot with the current registry to find all the crap it left laying about. The idea behind this being that we could do this with every app we installed and therefore be able to clean up after installs. Then somebody had an idea, lets see if it leaves any files around. Same again and the test applications tended to create files in the windows directory and shared directories and the root in some cases and then just leave them there when they were uninstalled. Finally we looked for fire modifications and yes, installers created entries in various config files and then failed to remove them on uninstall.* Suddenly the task of creating our home brew registry cleaner for every app likely to be installed on the network started getting a little scary.
Buuuuut.....
You still have to blame MS for most of this. They didn't have proper published standards and what they did have was not enforced.
*Something that blew my mind when the registry came along was that multiple windows config files still existed. Why could they not be merged with the registry? And things like the start menu folder were the same, why did that still exist if it could be merged with the registry? Answer, because nobody was taking overall control of design and development.
IIRC Novell's Zenworks produced a manifest of everything that an installer changed during installation, which slotted in perfectly with Novell's understanding of scaling enterprise networks.
If Novell had been the dominant force in computing there would arguably be a tiny fraction of the problems that exist today.
Most developers are more interested in writing installers than uninstallers
Most developers would rather be working on the thing that is to be installed, not the unglamorous installer. That job - and the uninstaller - gets left to the junior on the team, and he's not really interested either so doesn't take a great deal of care.
I mean, our application is great. Why would anyone ever want to uninstall it? This code will never be run ... there is no point in testing it!
Some registry bloat can cause extremely slow logons or the machine may fail to mount the system hive if you let it get big enough.
Example:
https://support.microsoft.com/en-us/topic/registry-bloat-causes-slow-logons-or-insufficient-system-resources-error-0x800705aa-in-windows-8-1-82a985fb-df27-abda-440b-f3f81a2c949d
Crappy printer drivers are another thing that will bloat your registry along with some applications.
As a terminal server admin, I have to keep an eye on the registry. Something that has little effect on a single user PC can have a big impact on RDS/Citrix servers. If I see NTUSER.DAT files getting much above 10MB, I will start investigating.
It's increasingly the case that applications don't have installers on Windows, and so they don't have uninstallers either. This is because apps like Steam etc.in the simplest case just dump a pile of files into a folder and expect it to work.
If those apps need registry entries then they will add them when they are first run.
There is no uninstaller, and so no way to then later remove them.
Wrong scarecrow, 30 years of registry problems in windows put the burden of proof on the other side argument. And the OP didn't mention registry bloat, that's your straw-man.
It has gotten "better" over the years, but there is still plenty of hot garbage in the registry. Last I checked it still allowed storing Unicode entries, but only displayed non-Unicode keys, allowing hidden entries(as leveraged by common windows copy protection schemes), helps enable plugin-hell and a death by a thousand cuts as a machine accumulates "run at startup" cruft that may or may not uninstall itself when the spawning app is removed.
And every crap Adobe app is still allowed to re-infest your boot experience even if you only run their apps once a month. Still no user controlled kill switch to keep apps from re-loading their crapware against the owners wishes without using 3rd party software.
Yeah, the windows registry is fine.
> If I get a new appliance to replace an old one (ie, upgrade an application), why should the fact that my basement is full of crap be an issue at all to said application?
If said application is a Roomba, it's not gonna be happy trying to cleanup navigating around all that junk.
This post has been deleted by its author
Yeah that's why Windows needs a clean install once in a while.
You must be doing something wrong.
The computer my company supplies me with is nearly three years old and despite having multiple versions of Visual Studio, multiple 3rd party SDKs and other software developer cruft installed on it is just fine.
The laptop I'm typing this on is still running the original version of Windows 10 it came with and it's nearly four years old.
My email/TVersity/Logitech Music Server machine is still running Windows 7 (although I'm intending to finally replace it with new hardware running Win10 this winter), it's not had to have a re-install either.
I usually re-install Windows when 1) I buy a new PC 2) a new Windows version has to be installed. That means it could be years between a a reinstall. The PC I'm writing this was installed with 7 when it was bought and then installed with 10 when 7 went out of support. Non need to reinstall it since then.
But this PC is a "stable" machine that gets installed only the software I need, and its updates - and they all original software - no "second hand" installers, or other "unsupported" ways to install software. Today, if I have to try some software before buying, it's usually run in a VM or in the Windows Sandbox.
The only time I had to clean up the registry myself was when I replaced a Logitech camera. For reason Logitech should explain, it installed a filter driver into the USB and audio stack, and didn't remove it correctly when uninstalled, creating issues. The developers should have been punished with a cat-o-nine.
>But this PC is a "stable" machine that gets installed only the software I need, and its updates - and they all original software - no "second hand" installers, or other "unsupported" ways to install software.
Not experienced the joys of HP printer drivers...
I remember having problems with printing (W7) and the recommended solution was to uninstall and reinstall, only problem the uninstaller didn't recognise stuff, because it hadn't been updated... Try and install the new driver and it would try (and fail) to uninstall the old driver, only solution was a deep dive into the registry and manual deletion of keys, followed by file and folder deletions. I think HP finally solved the problem by shipping a clean-up utility that you had to run three times to scrub all traces of HP printers from a system.
There's a reason why all these so-called registry cleaners are so widespread - there's a market for it, because issues with the registry can and do cause problems.
The existence of a market for a product doesn't require that it actually solve a problem, merely that someone can make money. Registry cleaners offer to "help" with vague unquantified problems like "computer running slow" and "errors and crashes." That's a classic salespitch for fake remedies.
I've yet to see a credible explanation how unused registry entries could significantly affect computer performance. Incorrect registry entries, sure. Damaged registry database, sure. But extra ones?
"I've yet to see a credible explanation how unused registry entries could significantly affect computer performance. Incorrect registry entries, sure. Damaged registry database, sure. But extra ones?"
Try these:
https://support.microsoft.com/en-us/topic/registry-bloat-causes-slow-logons-or-insufficient-system-resources-error-0x800705aa-in-windows-8-1-82a985fb-df27-abda-440b-f3f81a2c949d
https://carlwebster.com/the-curious-case-of-the-bloated-default-profile/
https://www.neteye-blog.com/2018/11/microsoft-remote-desktop-services-2016-issue/
by the time I jumped ship to Linux on the desktop in about 1997-1998(at home anyway) I was reinstalling windows about twice a year(perhaps more). First Win95, then I jumped to NT 3.51(hoping for more stability), then NT4. I was super frustrated with the lack of stability on windows at the time.
My first paying job in 1998 was helping develop "embedded" computer systems for video surveillance that ran proprietary software(built on VB3 I think??) that required Win98, so I had my fair share of time messing with that as well, though only at work. I wasn't doing any software development, it was more of a integration position, I worked on the software end(in collaboration with the software vendor), and another guy worked on the hardware, then the company built custom chassis for them. Years later after I left my co-worker told me every single system they sold was returned at least once for some failure/problem. That was a good laugh.
XP and later windows seemed to last longer in my experience but I haven't seriously used windows as a daily driver anyway since about 2005(and that was only on a work laptop, had XP on it, and replaced the shell with "Litestep", and used a lot of cygwin - the IT staff could never figure out how to use my system). I certainly have had Windows 7 systems that lasted longer than a decade and did not need reinstalling however they were also never intensely used.
> if you're doing a fresh install of Windows every few years.
Now I tend to do one very 6 moths. Back then a lot more.
It scares me when people are afraid of a reinstall and would rather have software make it worse.
A reinstall is a lot easier than uninastalling the shit you have installed, and yes I include myself thatg does not install a lot. Faster metghod of new install is faster. Of course things I need to keep are on other drives (mostly) and the ssd games are backed up so magically appear to the installers without downloading.
Yep. It may or may not need cleaning at times but plenty of cruft ends up in there, and registry fixers such as various driver clean utilities are absolutely required due to sub standard installation programs that can't cope with all suitable configurations.
but the registry doesn't NEED cleaning. This machine that I am using right now started life as a windows 7 machine. It has had inplace upgrades from 7 -> 8.1 -> various 10. Not only that but it has been through 3 intel motherboards, each time ive backed up, changed motherboard, restored backup as bare metal.
It simply doesn't fixing or third party utilities. It just works. If windows starts to bog down, create another user or clear the profile. Thats about it really.
Sure, windows 95 was a little flakier but 2k and XP were stalwarts (lets pretend vista didnt exist), but these were on an AMD platform, I wasnt too keep on inplace upgrading and swapping to an intel platform.
It reminds me about a ASM project in college where we had to write an in-memory drive.
A colleague decided to add some compression for extra marks and his 1Mb (physical memory) drive ended up showing something like 1700Mb of available space
That kind of compression rate should have won him a Nobel or Turing award, at the very least :)
I remember back in the day when I used Stacker on a 40MB drive under DOS. It reported the capacity based on a highly optimistic compression ratio, a capacity that suddenly reduced when installing games that were already compressed to the last byte.
It ultimately did work, it was just a bit optimistic.
I was wondering when someone would mention the drive compression utilities (Stacker, DoubleSpace, DriveSpace, etc.)
Cute little things that ended up causing a lot of grief in the long run.
(the idea, for those that either weren't born yet, don't remember, or have blocked it out of memory, was that the machine had the bootblocks, The compression driver, and a great big whopping file that was the compressed version of the rest of the drive. Recovering files out of a compressed drive from a file system corruption event was one of the reasons why I drink now.)
NTFS compression is still available on Win10 (and probably 11 too). Right-click on a file/folder and go into it's properties, click the 'Advanced' button, and select 'Compress contents to save disk space'.
Given how fast modern CPUs are, decompression is practically instantaneous these days, so it might still provide a speed-up on slow storage.
On some really problematic systems, running CCleaner's disk clean after Disk Cleanup is helpful, given it also cleans various application caches. Not quite as good as early editions of Evidence Eliminator, but good enough to erase corrupt caches.
I've yet to discover a similar tool for MacOS - a friend is unable to upgrade to Big Sur due their system having 70GB of application caches that MacOS gives no simple way to purge.
The compressed buffer consumes RAM itself. That means page faults spike much sooner than before. The only benefit is potentially reducing swap hits IF the data compressed.
Sometimes it's a win, sometimes it's a loss. The part I dislike is that it's hard to know which way things are going.
They probably lost the will to remove it after the current maintainers removed support for some of the more foolhardy options from it's earlier versions.
Let's face it, "Delete Duplicates" was basically a kamikaze run. I'm glad it's gone. But the more recent versions have helped tame the winsxs hive and other "quirky" bits of windows that would swell over time and threaten the smaller drives that many endpoints shipped with.
I would hope the maintainers had more sense than the reformers...
you only need to compare the clarity and information density of the Disk Clean-up dialog box with the Storage Sense settings menu to see stupidity at work.
My display is HD not CGA, yet the original MS Windows UI designers managed to clearly get more information across than many of the modern UI designers, with more pixels and screen inches at their disposal, could imagine.
OH GODS NOT THE WINSXS FOLDER...
(You know it's going to be a rough day with the winsxs folder is the largest folder on the entire machine; doubly so when the machine is a server that's running a LOB app and has run out of spare space for updates to be installed. TRIPLY so when it's a physical box and you can just throw another 20/40/100 GB at the boot drive because it's just not possible on a physical box...)
Ahem. Nerd alert.
RAMDoubler for MacOS functionality had nothing to do with MacOS’s pre-emptive multitasking capabilities (or lack thereof) and everything to do with the odd way in which RAM was allocated in Classic Mac OS. In fact, to be totally accurate, MacOS 8.6 and newer did have some limited pre-emptive abilities for Carbon applications, but still retaining the odd memory allocation scheme.
The odd memory allocation scheme was that it was manual. The developer of the software specified the memory range to be allocated in advance (overridable by the user) and the application took as much unfragmented memory as it could up to the maximum value specified by the developer. If it couldn’t get the minimum specified then it wouldn’t launch.
The upshot of this was that programs might have oodles of unused RAM but that RAM couldn’t be freed for use by something else. Additionally, you could have sufficient free memory, but fragmented, and the application still wouldn’t launch.
Another side effect of this was that virtual memory, as implemented by classic MacOS was spectacularly inefficient. And slow.
Connectix RAMDoubler on the Mac worked by a) defragmenting memory on the fly, b) allowing unused memory allocated for one program to be used by another and c) implementing an efficient virtual memory system. In other words, it just did what Apple should have implemented in the first place. RAM wasn’t doubled - but it wasn’t wasted either.
/nerd-mode
I suspect the unwarranted linkage to preemptive multitasking may relate to the reason than Chen got involved at all — per his blog entry he was chasing up on crash reports, and SoftRAM not only didn't actually compress but also was largely based on out-of-date Windows 3.1 DDK sample code, which being for Windows 3.1 made no effort to be thread safe.
So SoftRAM would crash hard on Windows 95 as soon as a lot of processes started hitting memory issues at once.
VM before around MacOS 7.5 was totally inefficient because they had a bug where writing out consecutive pages marked all but the last page as dirty again. It usually tried to write dirty pages in blocks of 8, so 7/8ths of the pages came back dirty and were swapped out again.
At the time I was working on a quite memory hungry application that had very bad launch times, so I had to read data at program launch in a strange pattern that minimised the effect of this bug.
And now the general opinion is that very, very fast memory compression is what allows macs with 8 GB of RAM to work like older Macs with 16GB. That and having a very fast SSD.
Windows registry does sometimes need cleaning, but in a forensic way. Sometimes things will bloat keys (dodgy printer drivers, shitty software). Let it get bad enough and logons will slow to a crawl or the machine may not even be able to mount the registry hive.
You can use tools like sysinternals RU (https://docs.microsoft.com/en-us/sysinternals/downloads/ru) to analyse the registry and compress it after you have cleaned it up. In a business environment, I have often had to use powershell scripts calling RU to clean up multiple user profiles to cope with bloating.
But yes, avoid "reg cleaner" apps like the plague.
People don't understand about speeding up computers much. On the one hand, I've lost count of the number of people who say that they've put their really important photos on an external hard disk because their computer had "too many files on it". Too many files meant "nearly 2000 photos" and of course putting them on the external mains-driven hard disk just meant the data was more likely to be lost than if it was kept on the computer (that was backed up). All that had to happen was someone tripping over the mains lead and yanking the thing off the table. "but I put them on the backup drive because it's a backup, and deleted them from the computer because it's a lot of files and they were clogging it up".
And on the other hand there are the people who run a defragmenter to "free up some space" even though it's an SSD and it doesn't free up space.
What's amazing is that it *could* have worked if they'd written it. The code was there. I don't actually recall this much. I vaguely remember seeing something about doubling RAM and thought that it's highly suspicious because you can't actually double RAM. I was partially wrong, but I'm glad I didn't fall for it.
I remember being partly impressed and partly disappointed with the doublespace type programs that increased hard disk space.
On another note, I read the Dr Dobbs article linked to, and I'm still impressed that they're talking about 4 and 8 MB machines, and now we're talking about 8 and 16GB machines.
@anthonyhegedus
.. But only to say that they are small amounts of RAM in 2021.
I've seen plenty of machines with far more RAM in them in recent years - servers with 256Gb or more of RAM and workstations with 64Gb or so of RAM and a couple of IBM mid-range systems that could scale to multiple TB of RAM if you had deep enough pockets.
Heck you can even get USB memory stick with 1Tb or more of storage in them.
This is not a "mine is bigger than yours", just an observation that we are already well past the numbers you are referring to.
Yes I was just talking about the “typical” system, such as it is. The typical system in the 1995 had 4 or 8 MB. I think typical home user systems and typical office systems are around the 4-16GB range now. My pc is 32gb. I’m aware that many modern systems are much higher. I remember in 1991 or 1992 we took delivery of a 256MB Sun Sparcstation. I remarked at the time that it was such a HUUUGE amount of memory, I couldn’t think what could fill it up. I think we were working with 4MB sparcstations at the time (can't remember!). I genuinely thought that you'd have trouble using all that RAM.
I was a programmer then.
Doesn't one of the Windows Update issues fixes involve cleaning up a portion of the registry? Also deleting files from a folder after stopping a service or three. And proper incantation in Creole while holding a chicken foot in iguana urine and poking the computer with a segment of black locust root.
You have to love the superior closed source tool set, no doubt gzip was deemed far too useful by some at the time (or more likely not ported yet). I barely remember this fiasco, but then recollection is probably lost in the swirl of other similar software versions of the chocolate teapot.
I remember some magazine review -- don't know which -- that literally concluded "SoftRAM 95 does what it says" (so there was a bit of a kerfluffle when it came out this software did nothing, since it also indicated this magazine was not actually testing utilities before giving them the thumbs up.)
As for Linux's stuff for this -- it's crap. zram is not a swap cache, it's swaps into a compressed ram disk (so some android tweaks do turn this on, since it allows gaining "some" ram without swapping, which should not be done to mmc or sdcard.) zswap? it's a compressed swap cache, but only up to 2:1 (with zbud) or 3:1 (a modified zbud that at least allows 3:1) because that's how it stores compressed pages (so a page "can" compress 50:1 but will still take 1/3rd of a page.) Worse, there's no page eviction!!! (Least recently used zswap is not swapped to disk when the zswap is full.) So your first swap is compressed and stays in the zswap cache, it's NEVER swapped out, so you then have a full zswap cache and any additional swap going straight to disk (disk swap is not compressed by this either.)
Quote: "A solid-state drive (SSD) doesn't need defragging, but Windows does it anyway."
Nope, can't comment for older Windows, but certainly any current Windows (i.e. 10+) does not 'defrag' SSDs.
The disk optimiser in Win 10 simply runs trim (once a week by default) if it's an SSD, it will only defrag if it's a HDD.
