[naviserver-devel] ns_adp_include -cache $ttl

[Stephen D. <sd...@gm...>]

ns_adp_include takes the new -cache and -nocache options. -nocache is
a boolean which suppresses caching. -cache takes an integer number of
seconds which is the amount of time the result of evaluating the ADP
code should be cached.

I think a better name for -cache would be -expires. We already use
this terminology for ns_cache_eval and friends. -cache and -nocache
look like two opposite boolean states, but cache actually takes an
argument of seconds.

(Hmm, do we need -nocache? Should -expires 0 mean 'no-cache', expires
immediately, or does that look like 'never-expires'?)

Currently you pass a TTL to -cache, the time to live in seconds.
-expires should support that. But it should also accept an absolute
time in the future for consistency with -timeout etc., the semantics
of which we've discussed in the past.


ns_register_adp and ns_register_tcl also take a -cache option, which
should also be changed (I added these, taking the lead from
ns_adp_include). Interestingly AOLserver 4.5 has changed the config
file syntax for marking which pages should be parsed as ADP:

  ns_section "ns/server/server1/adp"
  ns_param map [list /yada/*.adp   1200]

The page can now be a two element list with the second element being a
ttl. With the -cache option to ns_register_adp (which AOLserver
doesn't have) this config style can be neatly handled here:

http://naviserver.cvs.sourceforge.net/naviserver/naviserver/tcl/config.tcl?revision=1.2&view=markup#l_123

But here's the question: would it be better to add -expires to
ns_limits? It already handles -timeout.

http://www.crystalballinc.com/vlad/software/naviserver/files/mann/ns_limits.html

I wasn't sure at first but it's making more sense the more I think
about it. The -expires limit would be a hint to any command which has
some caching ability to, if not explicitly given a value, use the
expiry from the per-url limit.

Does this make sense?

I ask this now because it changes API. For the future, it might be
nice (and seems easy enough) to also add HTTP caching headers to the
output if an expiry is given. So, not only do we output cache, but the
browser won't bother sending if-modified-since requests.