Karl Fogel <kfogel@newton.ch.collab.net> writes:
> - On commit (to guarantee that the svndiff data sent is calculated
> against the right data)
>
> - On checkout/update (to guarantee that the svndiff received gets
> applied against the right data)
>
> The commit case is an entirely client-side change, and imho should be
> considered part of issue #549's milestone. The checkout/update case
> should not, on the other hand: it requires changes to server, client,
> and protocol, and anyway should be done at the same time as issue #649
> (fs checksums), so the server doesn't recompute checksums every time
> it updates a client.
Sorry, I should have said that more clearishly...
Of course checkouts will compute and store checksums locally as part
of #549. That way they're available to be compared against at commit
time. :-) And updates will do the same. The part that happens later
is receiving a checksum from the server and making sure it matches the
local checksum.
IOW, the first step is protecting against local corruptions either due
to someone accidentally editing a text-base, or (more likely) our
eol/keyword translation routines messing something up.
Using checksums as a sanity check w/ the server is the second step.
There, I think I actually said what I meant this time.
-K
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Apr 12 03:27:40 2002