Re: vcdelta, xdelta, or what?

[Karl Fogel <kf...@galois.collab.net>]

"Terrel H. Shumway" <ts...@octavian.ICS.UCI.EDU> writes:
> I noticed that the VCDiff internet draft is due to expire (tomorrow), 
> and I cannot find any references to an update. It seems to be a dead
> "standard".
> 
> way back in June, Ben Collins-Sussman wrote:
> >  We've been in discussions with Josh MacDonald (the author of xdelta)
> > for the last week.... more soon on this.  :)
> 
> What were the results of the discussions with Josh?  

Not definitive.  He appears to be busy with other stuff.  Branko is
implementing the vcdelta algorithm.  Nothing in Subversion really
requires that vcdiff be the transmission format for text delta
windows; if Branko thinks some other format is better, maybe we'll go
with that... (i.e., any reasonable format is probably good enough, it
doesn't have to call itself "vcdiff" -- but vcdiff is one format that
looks good).

I don't think an Internet Draft's expiring makes it any more dead (or
alive) than it was before.  If people use it, then it's a standard; if
they don't, then it's not.

> If we are going to used xdelta instead of vcdelta (I think we should),
> then we should probably remove the reference to vcdelta from the TASKS file.

The work required to use xdelta seems equivalent or greater to that of
just implementing vcdelta (at least when I was examining the issue in
detail a while back), and Branko has already started on vcdelta stuff,
so I don't see a reason to change...

Unless you have, or will soon have, a working standalone xdelta
library -- that would change everything! :-)

Re: vcdelta, xdelta, or what?

[Karl Fogel <kf...@galois.collab.net>]

Greg Stein <gs...@lyra.org> writes:
> However, the SVN FS will still need to record the type/version of the
> diff-encoder when it performs its "lazy diffing". That type/version would be
> transferred to the client along with the data; the client can then determine
> whether it actually has a decoder for that diff type.

Recording the diff style in the FS has been planned since the
beginning, yeah.

Re: vcdelta, xdelta, or what?

[Greg Stein <gs...@lyra.org>]

On Sat, Sep 09, 2000 at 03:37:11PM -0400, Jonathan S. Shapiro wrote:
> A small suggestion:
> 
> Within the SVN delta encoding, have a word that is used to indicate the
> format of the encoding. Also, build all clients from day one to complain
> about unrecognized encodings. We can then choose a different encoding at
> some future time, whatever the outcome is now.

At least from the DAV perspective, we can use MIME types, Content Encodings,
etc to signal this information.

However, the SVN FS will still need to record the type/version of the
diff-encoder when it performs its "lazy diffing". That type/version would be
transferred to the client along with the data; the client can then determine
whether it actually has a decoder for that diff type.

Cheers,
-g

-- 
Greg Stein, http://www.lyra.org/

Re: vcdelta, xdelta, or what?

["Jonathan S. Shapiro" <sh...@eros-os.org>]

A small suggestion:

Within the SVN delta encoding, have a word that is used to indicate the
format of the encoding. Also, build all clients from day one to complain
about unrecognized encodings. We can then choose a different encoding at
some future time, whatever the outcome is now.