Re: Better pdiff handling for apt
* Anthony Towns:
> Some time ago (*cough* 2009), I had a play with working out how to
> apply pdiffs more efficiently than apt currently does, and implemented
> a proof of concept in python [0]. There weren't any replies (even a
> "ooo, cool") when I posted to the deity list, so I left it at that;
> though trolling google now, I see that it got mentioned on -devel [1]
> not all that long ago (*cough* 2012), so apparently it did get read
> after all. Not that it looks like it would've done much good, because
> it seems like the script I put on the web, and the one I was actually
> using are completely different to the extent that the one on the web
> even has syntax errors. WTF?
I think there was also a proposal to merge the diffs on the server
side, so that clients would only have to download a single diff file,
as identified by the starting checksum. I've never seen a
specification of the file format, so I don't know if that would be
backwards-compatible.
And in the past, I think the APT method API was not considered
sufficiently flexible to support such merging, so that why I stopped
working on the patch file merger I had.
But that is purely from memory, I have trouble finding actual traces
of previous discussions.
Reply to: