Bram Cohen: The diff problem has been solved

bramcohen.livejournal.com

18 points·aston··6 comments

I can vouch that BZR has one of the best plain-text diff tools - whether or not it's number 1 without numbers to back the claim is a different matter though.

However, what I really want to see is BZR and GIT catching up to SVN when it comes to binary diffs - SVN remains the most efficient at finding and commiting changes to binary data between revisions. It may not seem so important, but if you're ever in a backwards country on business and need to upload to a repository via dial-up..... and of course, the smaller the binary diffs the more space saved on the server and bandwidth charges.

What kind of binary files do you put under SCM?

Mostly binary dependencies. Some are executable files (for instance, GRUB Stages 1 & 2), some are DLLs, and some are redistributables.

Reason they're under SCM is because we maintain some of them; with varying degrees of changes. Some have their own repositories, naturally, but others involve minor changes and/or just simply take too long to build. And for some of the redistributables: we don't have the source code for them in the first place.

Images and MS Office documents. If you do anything with MS Access then the program itself is stored in a binary file.

What exactly does it mean for this problem to be "solved"?

edit: Bram says "It's been solved in the handwavy 'we got it to stop giving crap output so often' sense."

Just noticed the date on this, I wonder if something a little more formal than a livejournal post has been written about it? If so I can't seem to find it, but maybe someone else has had better luck?

Reply on news.ycombinator.com