On Tue, 16 Aug 2005, [UTF-8] Branko Ä^Libej wrote:
> Peter N. Lundblad wrote:
>
> > And what does 2.0 have to do with this discussion?
> >We're not talking about moving the line-based diff to the server, as I
> >understand it.
> >
> >
> No. And if I'd thought about it a bit, it would be obvious that if the
> server is to compute deltas, it must compare file contents (at least in
> principle).
>
> But won't the client have to actually download the delta in order to
> figure out if the file has actually been changed?
>
Not in the ignoer_ancestry case according to the current code. Then
svn_repos__compare_files, which will do an actual byte-by-byte comparison
after some shortcuts have been tried.
In the --notice-ancestry case, you'll get false positives. Then, the
apply_text_delta call tells you something like "the contents have changed
between these versions, but they might be the same anyway" or something.
For the summarized diff we're discussing here, it seems to make sense to
use ignore_ancestry, since that's the default for a diff with deltas. If
it seems necessary, it would not be hard to add another flag, like
no_empty_text_deltas or something. I'm not sure how this feature wil be
used most.
Thanks,
//Peter
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Aug 16 08:31:33 2005