[Seaside] Re: REST and Seaside

[Yanni Chiu]

Nevin Pratt wrote:
> 
> Comanche does not give detailed "hit" statistics like Apache can.  I
> have some code in there that gives me some hit statistics, but I
> disabled that code about 5 months ago, wanting to see if it improved
> anything by not giving Comanche as much to do.  It didn't really make
> any difference, but never-the-less I never re-enabled the hit code.  I
> figured that since every page has to also download pictures from machine
> #2, I would just use hit statistics from machine #2 to "monitor" things.

The stats may not be accurate because of caching of static
pages by the web browsers and by the ISPs. Once a user has
browsed your site, then many of the images/pages will already
by cached. When repeat customers come back, they probably
hit your Seaside box much more heavily than your static
page serving stats are indicating.

Here's a wild guess. Could the session timeout be suffering
from a timer rollover of some sort.

Another thing I noticed (when testing untuned code) was
that when I had an 8000 plus dataset, memory consumption
grew rapidly within a few clicks. What I think happened
was that the state holders were making shallow copies of
the collection. I seem to recall that Seaside cut the
chain off at about 10, but if each saved state is large,
it can still be a problem. Maybe a faster expiry for "large"
sessions could be useful.

Congrats on the success of your site, and I hope Seaside
continues to work for you. BTW, do you have any stats on
how hard Postgres is being hit.

--yanni