Couldn't it be as straight forward as doing a composite cache in the Jsp
Servlet-- an LRU of a specific size that demotes to a weakhashmap?
Leave the compiler as is (the whole thing should be re-written though).
Costin Manolache wrote:
On 3/5/06, Remy Maucherat <[EMAIL PROTECTED]> wrote:
Costin Manolache wrote:
On 3/5/06, Remy Maucherat <[EMAIL PROTECTED]> wrote:
I also don't see any need to have zillions of JSPs filled with template
text, especially if you're willing to take a small performance hit
(there's that thing called dynamic includes which could be used to
handle large portions of static text).
Well, unfortunately tomcat-dev's role is to implement a servlet
container,
not to
decide how people should use it and how many JSPs or other technologies
should
the use, or how they should set up their hardware :-)
I don't think it's good to target a single use case or set of users -
i.e.
huge servers, and
sites with small enough number of JSPs to fit in memory.
This is blatantly false. tomcat-dev's role is to come up with an
implementation. There's nowhere anything specified about complying with
anyone's needs. Here, I consider the needs of this user ridiculous, so I
will not even consider them.
Using less memory and supporting users who don't run tomcat on huge servers
is not
ridiculous. And 'this is not my use case' is not a valid reason to veto IMO.
Many cases would benefit from more control over memory - hosting or
embedded or sites with lots of jsps or lots of data. Forcing all
static content in memory is not the best use of the memory.
There's no other solution really. Any other implementation will perform
bad, due to the very fragmented nature of static text.
Apache seems to do ok with serving static text without loading it all in
memory :-)
And except JSP, I don't know many other templating solution that
requires
all data
to be in memory - of course, not everything is as fast as JSP, but raw
speed is not
the only concern :-)
You are trolling here. I would be more than happy to serve JSPs as
static text, but somehow I cannot. Again, JSPs are not a regular
templating solution since they have to be compiled. This is not all bad,
since this allows easy optimizations too (for example, the JSPs are GC
friendly).
How are JSPs GC friendly ??? By keeping all strings referenced by the class,
so they
can't be GC until the webapp is unloaded ?
I think it is perfectly possible to compile JSPs in a different way - the
current compilation
is not the only solution or only way to generate java from the jsps.
I also agree with you that JSPs are complex enough that messing with the
code generation
is too dangerous - I wouldn't touch it, but if some people have patches, we
should consider
them.
Costin
--
Jacob Hookom - Minneapolis
----------------------------
JSF-EG, JSF-RI, EL, Facelets
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
