Hacker News Reader

Comment by EgoIncarnate | Hacker News Reader

It seems Mark Pizzolato was at the time the primary maintainer and recent contributor (by a large margin) (https://groups.io/g/simh/message/1501).

Mark made a change to how the virtual disk system worked that broke an esoteric use case (using dd to image multiple disk images to an SD, without an actual file system). The user didn't want to use a workaround (specifying the transfer size in the dd command), and proceeded to use twitter to get people to harass Mark. (https://www.bentasker.co.uk/posts/blog/the-internet/toxicity...)

Mark got sick of the harrasment and decided to take his toys and go home.

Toxic user prompts toxic reaction by developer.

View on HN

I'm sure glad I don't work with developers like that, because that complaint makes perfectly reasonable sense to me and I certainly wouldn't consider it "toxic". Expecting a raw disk image file to be the exact size of the virtual device is not at all an "esoteric use case".

Being a prolific contributor doesn't mean you get to force your mistakes on others and cry about it when your decisions come back to bite you. See what Big Tech is doing to us for a great example of what this "feelings-driven design" crap has given us.

Words to live by: "if everyone tells you that you're wrong... maybe you are."

It sounds like there was a config option available to disable the signature addition to the image file ( https://github.com/simh/simh/issues/1059#issuecomment-108689... ). I could see a benefits for having an embedded image signature for preservation and corruption detection issues.

I don't think complaining about the design is toxic, but recruiting uninvolved people on twitter, and harassing out of band certainly is. Also reading the bug thread it seems the person with the issue wasn't the same as the one who instigated the harassment. (https://github.com/simh/simh/issues/1059#issuecomment-108675...)

Enabling it by default could break lots of valuable data over the history of computing. That's why I prefer simh-classic.

Not really; the signature is appended to the tail and can be safely ignored by things that don't know about it. This is a pretty thoughtful way to make the change.

It only really matters if you are doing something weird -- e.g. concatenating a lot of images together.

Having metadata instead of raw images is useful.

The thing was turned on for a couple years before anyone complained IIRC: and then suddenly it was a high-stakes issue with brigading, etc.

NEVER,

EVER

mangle your raw images, specially if you want to dd/tar them back to the original machine. Also, as a golden rule on IT, specially on emulation, NEVER do unexpected changes to the original media, be disk images, CD-roms, BIOS dumps or any other firmware.

If you want to make changes for interop, do it in RAM as a volatile change or write the metadata anything else on the emulator config directory.

The mentiones change is not useful, it really sucks. The move was one of the words turds ever done to historical computing and data preservation. Period.

The purpose of emulation is to run and preserve the original systems -unchanged- by design. The user is the one who commits the changes while he is running the emulated system from the inside, and not the emulator.

The goal is to preserve the data back and forth, not to ruin it.

> mangle your raw images, specially if you want to dd/tar them back to the original machine

dd back to the original media is no problem with stuff post-pended; dd will complain that there's an extra block at the end, but other than that, all's good.

> The goal is to preserve the data back and forth, not to ruin it.

A big part of the point of this change is to prevent you from attaching devices "wrong" in ways that are likely to cause the emulated environment to corrupt them. You can't do that in RAM, and it's better if it is in the same file.

>You can't do that in RAM, and it's better if it is in the same file.

No, that's a setting for the config file, or per machine by default in the settings, but never in the media itself.

The old media should be pristine. If the user mangles some media because it attached some cd image as an RP disk under a Vax, it's his issue, is not an issue of the emulator.

It's not "old media". It's a disk image in some format.

Before it was just a raw disk image.

Now it's a container format for a disk image--- that happens to generally work as a raw disk image, too.

Either one keeps the original data pristine.

No, you are wrong. Emulators should never touch the original data format save up for converting it at will, as qemu can do by using qemu-img.

That's a belief you have, but it isn't something that one can be "wrong" about.

That "esoteric usecase" was one reason people opposed that change. There are other concerns as well (you can find them if you read the discussion in the GitHub issue linked elsewhere in the comments).

For example, it made SimH alter disk images by itself, without warning, even if you only wanted to use them in a read-only manner.

There was also concern that sticking metadata right into the data itself was bad from a preservation standpoint.

Also there was the point that such things should be opt-in instead of just appearing ferom one version to another, without warning.

All of these concerns were dismissed as unimportant or flat out ignored by Mark. Then he went full overboard with his license change (which made a free project non-free)

Edit: removed incorrect reference to the GPL

> you're not allowed to add additional usage restrictions to GPL software

SIMH is not GPL'd software.

yeah, you're right, of course. What I meant is that it made a free license non-free and that got people concerned

Maybe they shouldn't have harassed him -- sounds like he was doing it on his own time.

Presumably still is -he continues to be involved in both projects (judging by the commit histories).

> an esoteric use case (using dd to image multiple disk images to an SD, without an actual file system).

fwiw this is an _extremely common_ use case. it's literally one of the tasks that dd was created to do

Calling the use case esoteric is flatly wrong. It's literally how many disk images across decades of software work. It's how things like vnd(4) work.

Nobody would've cared if it 1) wasn't turned on by default, and 2) didn't destructively change disk images.

It's a bit extreme to call people toxic because they got upset that a deliberate change caused data corruption.

I don't think getting upset is toxic. But when you don't get the response you want from a small open source project, your next steps shouldn't be to encourage more people to bother the developer(s); that's the toxic behavior IMHO.

Agree. I see the original post was making a distinction between people who were upset and the person who tried to encourage harassment. That person definitely was being toxic, so please disregard my comment about calling people toxic because I misread and agree with the label

Nobody cared for a couple of years until one person started recruiting lots of people to bug the developer.

It doesn't destructively change disk images; it appends something to the end. Other usages of disk images are fine: it's only if you want to use the disk images without a filesystem that you run into trouble.

Metadata is useful. Metadata is a whole lot less useful if it's off-by-default.

This is decidedly not the case, and it attempts to misattribute the issue.

I cared because it happened to me. Like most people, I use a package management system (pkgsrc), so I wasn't on the simh mailing list and I didn't know about this change. I update simh via pkgsrc, tried to restart some virtual machines and had fscks corrupt the filesystems of those VMs.

I went looking in the mailing lists, saw the discussions, saw how people asked nicely for him to do something more reasonable (the most reasonable would be to not default the feature to on), and saw how he simply doubled down and became dismissive.

Note that I don't advocate for harassing people at all. That's a separate issue, and I condemn anyone who'd encourage harassment.

"Metadata is a whole lot less useful if it's off-by-default." You're making excuses for him. We shouldn't force something on others because of specious suggestions that it might be useful in the future, particularly when it's actively harmful right now.

How does fsck corrupt filesystems when the metadata is something that's post-pended past the end of the image?

The reason why the original user (al20878) in the issue had a problem was because they were storing the images without a filesystem and expected them to have never changed size when they copied the image to the SD card. This isn't a typical use case.

Can you explain how you ended up with corruption based on a small thing being postpended to a raw disk image?

I honestly don't know with certainty. I explicitly set drive types with "SET RA* TYPE=" (and RQ*) and image sizes with RAUSER, and suspect that the metadata got written to after the drive type's size offset in to the image.

I noticed several other bugs: one of my disk images has the metadata block at the beginning of the image (no idea how). One of my systems (a big endian system - imagine Mark's shock and surprise when he learns that people who run less common systems in simh sometimes run... less common systems!) had appended an extra metadata block onto each image every time simh was relaunched.

Once I realized that there were these blocks of extraneous data here and there, I started looking online to see if others had similar experiences, and came across the threads talking about this new "feature".

And I disagree that writing disk images to disks / partitions without explicitly giving dd the size isn't a typical use case. I can't remember the last time I specified the size. The size of the file is always assumed to be the size of the write, and it'd be a very special edge case to deliver a file, then tell people to only write a specific size. Go download any .iso or .img file for installing any OS out there, and show me on the download page where we're told to stop after a certain byte (block) count if you disagree.

> I started looking online to see if others had similar experiences, and came across the threads talking about this new "feature".

OK, then it sounds like you have a legitimate bug: a far clearer report than anyone else has provided here. Perhaps you should communicate it.

> And I disagree that writing disk images to disks / partitions without explicitly giving dd the size isn't a typical use case. '

I didn't say that. I just said (in other comments) that if you overrun the size of the device or partition, the kernel reports an error and then dd whines that it couldn't write the last bit.

  $ sudo dd if=/dev/urandom bs=512 of=/dev/fd0
  dd: error writing '/dev/fd0': No space left on device
  2881+0 records in
  2880+0 records out
  1474560 bytes (1.5 MB, 1.4 MiB) copied

The only person in the issue with data loss is the reporter, who was dding image files onto offsets on an unformatted SD card. When one of them got longer, it created problems-- overwriting the first bit of the next image. If you're doing something like that, you really ought to be providing the size.

Reply on news.ycombinator.com