[REBOL] Re: Rugby performace starting to fade
From: petr:krenzelok:trz:cz at: 19-Feb-2002 14:20
Oh well, having 6 hours of sleep in two days (watching Czech ice hockey team in
Salt Lake City ;-) I am really not prepared to hear something like that. Your
reply suggest that I try to say something bad about Rugby intentionally, while
just opposite is true. You skipped rest of my request, in which I ask someone else
to confirm my results.
In fact - I do care, as if I wouldn't, I would never try to download Rugby. The
problem lays somewhere in between - if I want to know something about Rugby, I
have to study source code, or just ask. Source code is being changed, so it is not
all that easy to keep track of all the details.
So what did I actually do when testing? I simply started 4 rebol consoles, all
from one and the same directory, using the same config file! I ran one example
using old Rugby, one example using new one, with results I posted. I got results I
got, so what? My testing procedure seemed to be logical to me.
Of course, after reading your response I gave it another try. Thre results dropped
to nearly one minute, so I actually looked into proxy stuff. My user.r now holds
one line setting system/schemes/default/proxy/bypass block, containing two of our
company servers, as well as "127.0.0.1". But it also contains my local IP address.
So I just added new IP address of machine I was performing tests on to the above
mentioned bypass list, and the performance was back to half a second. I don't know
why "127.0.0.1" is not ok anymore ....
But - old Rugby was OK, while new versions require me to add each machine I run
test on to the bypass block. I think that there is difference in Rugby 4.3 httpr
or tunnel code and the one from current or XPi version. So, actually, where is the
truth? Is it my bad config? Or your change in the code? How should I know?
So, if you prefer only positive reports on ml which is here for developers to have
some feedback, more than actual, maybe even incorrect test results, I can stop
testing at all ...
I don't know what exactly "sigh" means in your other response, but one conclusion
- W2K code is OK, W9x is not. It was confirmed by Cyphre and Pat. Holger replied
that it is maybe due Rugby's architecture. I can understand him, as he has not
probably studied Rugby nor does he have enough free time to do so. He suggested to
look at tcpdump data, but I have not found any free time to do such deeper
investigation yet. Now I am not even sure I should care. I know where the problem
lies - simply don't use W9x with Rugby and communication intensive stuff. W2K is
OK. That's acceptable knowledge to me. You can put it in your doc, or you can
forget reported data.
OK, now I feel better ;-)
-pekr-
Maarten Koopmans wrote:
