> but  about  the  coparison  between PHP's autoload and CS-JS, i must
> disagree.  on PHP, the overhead is with milliseconds, and can easily
> be  made efficiant through caching. On CS-JS, no matter what you do,
> you  will  still  be doing more calls to the server (by default this
> would  be  for  every  missing  class),

Not  necessarily  to the server... with proper caching, only the first
request  would  be to the server, subsequent ones from the local cache
even  after  browser restart. And taking into account single-page rich
web  apps,  there  are  a  lot  of  cases where there is only one HTTP
request needed.

It'd be foolish for me to argue that multiple HTTP requests, even once
per  sesession,  won't  be much, much slower than loading .class files
from  the  server-side  filesystem. But that wasn't ever debatable. My
point  is  that  the  only reason to embark on an __autoload emulation
(which at least 4 of us are still into!) in client-side JS would be if
you  have  already  have  made the decision to use a file-per-class or
file-per-package  architecture  -- just as the PHP feature is built to
accomodate  an  existing  programming tradition. So assume you already
will  be incurring overhead when you load the classes. The question is
whether  building  a transparent autoloader can be done well enough --
adding  little enough overhead/pseudo-keywords/stylistic changes -- to
be superior to manually listing and loading dependencies.

-- S.



-- 
Subscription settings: 
http://groups.google.com/group/mootools-users/subscribe?hl=en

Reply via email to