[Zope3-dev] Asyncore (and Twisted)

[Leonardo Rochael Almeida]

On Fri, 2002-11-01 at 12:40, Shane Hathaway wrote:
> Chris Withers wrote:
> > Shane Hathaway wrote:
> > 
> >>
> >> Well, I don't think networking from threads is the real goal.  What's 
> >> important is that Zope's ZODB connections are scarce and precious, so 
> >> time spent in application code must be minimized.  To help achieve 
> >> this, Zope threads must never block for I/O except in rare cases.  
> >> Therefore all input and output must be fully buffered.
> > 
> > 
> > How does this interact with the (admittedly edge) case of pages where 
> > you want to write the output bit by bit?
> > (I'm thinking things like the dots that ZUnit used to write out...)

Wow, someone actually remember those :-) (Hi Albertaas, enjoying being
the new mantainer? :-)

> In cases like that, without the optimization, the thread must block 
> waiting for the main loop, which in turn waits for I/O.  With the 
> optimization, the thread can avoid the main loop and do I/O itself.  But 
> like you said, this is an edge case.

I, for one, have written a lot of streamming pages:

* ZUnit dots :-)

* Newsletter sending systems

* Full scale recataloging of objects (in situations where ZCatalog's own
"Find Objects" was so slow as to cause time-outs), and other
ZODB-rovering activities.

Basically any place there was a possibility of an activity taking too
long and we couldn't afford the browser to time out (clients wheren't
too keen on browsers timing out for some reason :-)

Admitedly, most of the uses of a streamming page are stressing the
limits of Zope apropriatedness (sp?) (i.e. few would argue Zope as the
best infrastructure for a Web chat app), but it's one of those cases of
"when you need it you need it badly".

So I humbly ask for some possibility of streamming pages, even if in a
suboptimal way.

	Cheers, Leo

-- 
Ideas don't stay in some minds very long because they don't like
solitary confinement.