Ramiro Diaz Trepat wrote:

>On 4/14/05, Tim Rowledge <tim at sumeru.stanford.edu> wrote:
>Ramiro Diaz Trepat <ramirodt at gmail.com> wrote:
>
>  
>
>>>I believe that no serious server software will ever be possible
>>>without preemtive multitasking, probably no client sofware either.
>>>      
>>>
>>Really? Have you not read about SeaSide, possibly the best webserver around?
>>And we have a variety of net clients for things like IRC, mail, ftp, a
>>webbrowser etc.
>>    
>>
>Are you serious ????
>I appreciate A LOT all the improvements in Squeak applications, and
>the servers like SeaSide, but if you really think that they can
>compete (any of them) with their non Squeak counterparts in the real
>world, you have to be the most irrational guy I red in years.
>Why do you think that no large web site uses Squeak?  Just because
>they are stupid computer illiterates?  The guys at Amazon, e-Bay,
>Google, Yahoo; they are just jerks, right?
>  
>

Now things are getting carried away.  They systems are relatively
unique.  They achieve massive data throughput by throwing lots of
distinct servers at the problem...nothing to do with multithreading. 
The technology they used is dictated by the fact that they can support
absoltely no session state (Seaside is out) and page generation needs to
be faster than disk access (they can't even serve static files in many
cases).  If that's your ball game then Squeak+Seaside isn't the tool for
you.  Frankly I wouldn't want to write web applications with the
restrictions faced by the Yahoo developers.  However, the original
poster is right IMHO, Seaside is one of the best web servers around for
the group of applications for which it is designed.  I certainly hope we
make every effort to improve performance but threading just isn't at the
top of the list (JIT + PIC being first on my mind).

>I know that languages like Java managed to do it properly, so I don't
>believe that argument.   Application Servers that are completely built
>in Java run perfectly on Windows, Linux, Solaris, etc. I see this
>every day.
>So there has to be a way.
>  
>

OK, now I know you're trolling.  Anyone who says any java application
server "runs perfectly" has never administrated one.  I've got
Tomcat/JBoss/apache and several years with a standalone WebSphere server
and I can tell you that stability is not their strong suit.  What I
can't say, and what you seem to be avoiding pinpointing in your
discussion, is if threading has anything to do with the problem.  One
can construct a Java servlet that uses multithreading and behaves
differently on a Windows machine than a Solaris machine.  Claiming that
the app servers "run perfectly" has nothing to do with portable
threading semantics.  It just means they've done there jobs and avoided
the pitfalls...again, in my opinion it also means that you've never used
them seriously (under memory and CPU load, that is).

David