“Slack transcodes some files you upload, including removing byte order marks ” | Hacker News Reader

“Slack transcodes some files you upload, including removing byte order marks ”

twitter.com

27 pointsobi1kenobi11 comments

> if your program accepts file uploads it should convey them byte-for-byte to recipients, good lord

You'd be surprised how few apps abide by this simple and reasonable rule.

My own Mattermost instance refused to allow me to upload an image to my own machine the other day because the long edge was 9000 pixels. "Too high res for your own hard drive" was a new one for me.

> if your program accepts file uploads it should convey them byte-for-byte to recipients, good lord

There are several good reasons not to do this; stripping EXIF data from uploaded images is a good one.

There's nothing wrong about stripping EXIF data, but you need to be upfront about this.

Haha, remember when FTP text-mode was a common preset? Bad times.

Way back when I first heard of this new thing called “Linux”, I downloaded all 3 CDs worth of RedHat 3, over a very long few days on a 33.6 dialup connection.

I then discovered the FTP client I used defaulted to text mode. Needless to say I was a bit annoyed at it, and then also myself for not looking into it.

I'm guessing on Windows it added 0A after every 0D?

I guess you could find all instances of 0D 0A, and then for each position, program the FTP client to download 2 bytes at byte $POS, and fix the file that way... Obviously you'll need to reduce 1 byte from all the next positions if there wasn't an 0A in the remote version.

Easier said than done, of course. I doubt I'd be able to do this at the age where RedHat 3 was around. And on Windows on top of that!

UTF-8, no BOM. Use it.

While I agree that utf8 should be used (whether Slack is involved or not), this is a blame-the-victim solution. Slack is utterly in the wrong here. And who knows what other files they are reading, much less damaging? Are they scanning your uploaded images? Doing ML or "Shazaming" your videos and monetizing the results like Roku and the TV mfrs are? Downscaling your images?

Perhaps they only do it to free accounts, where a benefit of paying is that your files are not molested? That would be both horrifying and yet, at the same time, hilarious.

Maybe I’m being idealistic, but wouldn’t a company with such a huge surface area of customers be wary of doing something like that intentionally? I couldn’t find any information about it in their Privacy Policy, besides a mention that they log basic metadata of files shared on the app.

This is the exact thing that wasn't working and which necessitated me sending various test files with particular encodings to a colleague.

It is seriously troubling that you think the default text encoding for nearly every contemporary text editor WASN'T tried.

Reply on news.ycombinator.com