[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Proposed editor interface changes for pipelining

From: Greg Hudson <ghudson_at_MIT.EDU>
Date: 2003-10-27 04:06:45 CET

I would like to document the following editor interface constraints:

  * If you get an error from an editing operation, you cannot assume
    that the error resulted from the particular edit operation you
    were trying to perform.

  * If you get an error from an editing operation, the edit is dead;
    you cannot do anything other than abort_edit().

All of our current code obeys this constraints. These changes allow
edit operations to be marshalled over the network without round-trip
delays, while still allowing errors to be reported back to halt the
operation.

I have ra_svn code ready to commit (passes all tests, against itself
and cross-versioned in both directions) which implements true
pipelining. Over my 10-15ms RTT cable-modem link, the new code
improves checkout speed by 100% (times measured with an export of the
svn tree, and fell from an average of 95 seconds to an average of 46
seconds.)

I believe these editor constraints are absolutely necessary for
pipelined operation, and have no practical downside. I'm going to
assume I have an implicit +1 from Tobias (who already proposed these
changes, but maybe not loudly and clearly enough for everyone to
notice), so let me know if you're -1 on this.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Oct 27 04:07:32 2003

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.