Following all the rumours about svn 1.1.0 performances I decided to make
further investigation on Windows->Windows interaction.
Unless there is something odd with the version I downloaded (most reasonable
explanation), I have the following results issuing a svn ls URL agains a
svn 1.1.0rc1 server (no mod_authz_svn troubles):
1) svn client, version 1.1.0 (r11180)
compiled Sep 30 2004, 03:04:31
Each PROPFIND (a total of 10) is split in 2 packets: the first containing HTTP 1.1
header, the second containing the xml DAV request body.
In between the two packets, client side receives an ACK packet from the server
with an approx delay of 200msec.
Since the total PROPFIND are 10 you end with at least 2 seconds elapsed time.
2) svn client, version 1.1.0 (r11180)
compiled Oct 4 2004, 21:11:18
Exactly the same situation.
3) Java client using svn, version 1.1.0 (r11180)
compiled Oct 4 2004, 21:11:18
Exactly the same situation. With the client making a new connection at each
"svn ls" equivalent command
4) TSVN 1.1.0
The entire interaction ends in 0.14 seconds (!).
The 2 packets in which a PROPFIND is split are sent in sequence without
delay.
I'll try to ask TSVN team what kind of magic they made.....
Guido.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Wed Oct 6 09:30:04 2004