On 7/2/03 2:01, "Niclas Hedhman" <[EMAIL PROTECTED]> wrote:
> On Thursday 06 February 2003 20:27, Pier Fumagalli wrote:
>> On 6/2/03 3:47 am, "Niclas Hedhman" <[EMAIL PROTECTED]> wrote:
>>
>>> Shouldn't you in this case have front-end load balancing routers (various
>>> topologies available) to lower the load on each server?
>>
>> Wish I could... I can't have (because of a fucked up design and
>> implementation before I came on board) 2 instances running (they clash on
>> the database side... Who is the idiot storing the status of a cache in the
>> servlet container in the database where the data to be cached is? Jules!)
>
> Hmmm. Maybe we should re-introduce the death-penalty for programmers....
Well, one way or another the guy is no more for us, soooo...
>> It's not thousands of dollars, boy I'm _not_ that expensive... :-)
>
> No?? $500 a day?
Gosh! I _wish_ :-) That would put me down around the 80k pounds/year...
Not even close! :-)
<note>
The first one who offers me a job thinking that I'm interested in money,
will be personally flamed by the underwritten author of this email...
</note>
>> Anyhow, the "who cares? Go and buy something off-the-shelf" attitude is
>> _so_ wrong... Ok, I'll go shopping in Tottenham Ct. Road and forget about
>> thinking that there might be a better and more intelligent way out???
>>
>> Gee, with that attitude noone would have even thought about writing Apache
>> 2, you have such a perfect solution going and buying M$IIS off the shelf
>> :-)
>
> Well, the idea of buying a solution to "load balancing" is to free up time to
> invent in areas where no solution at all exists.
> No matter how much you optimize a piece of software, you will always have a
> peak limit. Question would then be, how much time to spend for how much
> improvement, and until we reach which limit?
Well, the limit is "cleanliness"... Once a piece of software is "clean",
"linear", it is, at the same time, quite well optimized...
Probably my rants are not about absolute performance (which I can get out of
mere mod_rewrite hacks!) but on how one thing is "clean" and "linear"... I
have this feeling that easier it is, faster it is...
> Let me also say, that there is an enormous difference between I spending 2
> weeks to save $3000 of purchases (which I probably won't do, "why bother"),
> compare to me spending 2 weeks to save a whole community $100 each in
> purchases (which is a lot easier to do, "grateful people").
And if you spend 4 weekends (they come for free) to save a whole community
$3000 of purchases? :-)
> Anyway Pier, I am happy that you (and Stefano even more so) is expressing
> "concerns" (mildly) about Tomcat and its direction. I never really liked it
> (mostly stemmed from its early problems with configurations) and stuck with
> Apache+JServ (mostly out of laziness "It ain't broke").
That's why I resorted to Jetty, at the end... It's clean enough, it does its
dirty job of serving friggin' pages, and it's 1/10th of the size of Tomcat.
I also like ServletExec (same approach), but too bad that it's not open, and
sometimes I spend some time reverse-engineering it to fix some bugs! :-)
> Are we looking into a mod_cocoon somewhere down the line?? (That would be fun)
I don't believe that this community wants to see C code in the CVS
repository :-) :-)
The ideas are there, and a partial implementation as well... Right now it's
just too easy to use mod_proxy and let the baby live on its own...
One thing IMVHO we should focus on beforehands would be to have a "clean"
(minimal) Cocoon distribution... Removing the hard-coded "WEB-INF" paths and
write some more doccos...
Pier
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, email: [EMAIL PROTECTED]
