So the current scoreboard seems to be: 3.0s excess time to do this on
lpnet, vs about 450ms to do it on lp.dev.
It seems a lot of the difference between them is due to load or
queueing on the production instance. Since that's 85% of the problem
and since there is work underway to fix it, I think we should shelve
this until that is done, and then measure again and investigate the
causes of the delay vs plain bzr.
So the current scoreboard seems to be: 3.0s excess time to do this on
lpnet, vs about 450ms to do it on lp.dev.
It seems a lot of the difference between them is due to load or
queueing on the production instance. Since that's 85% of the problem
and since there is work underway to fix it, I think we should shelve
this until that is done, and then measure again and investigate the
causes of the delay vs plain bzr.
Martin