Re: lynx-dev automatic uncaching (was: PATCH [dev22]: source caching fixes)

[Leonid Pauzner]

16-Apr-99 04:53 Klaus Weide wrote:
> On Thu, 15 Apr 1999, Leonid Pauzner wrote:
>> 15-Apr-99 03:50 Klaus Weide wrote:
>> > If the behavior is changed from "only reload top document after
>> > relevant option change" to "reload all affected documents after option
>> > change", then please make this optional/configurable.  The current
>> > behavior is maybe sometimes unexpected, but I generally prefer it over
>> > the alternative.  I know where the ^R key is if I need it, and would
>> > like to avoid unnecessary reloading.
>>
>> I think a new behaviour should be implemented for SOURCE_CACHE!=NONE
>> (always "reload all affected documents" when we have sources locally -
>> it would be strange otherwise, so no configure options required here).
>> But with SOURCE_CACHE:NONE the old behaviour is preferrable by default
>> to avoid shocking of experienced lynx users base
>> (non default option may be implemented of cause).

> It would be nice if these two aspects - SOURCE_CACHE setting on the one hand,
> auto-invalidation of loaded documents on the other - could be implemented
> as independent settings (even if for the user they are coupled into one
> option as you sugggest).  That would make it easier to change things later.

Auto-invalidation logic controlled by a single line in mainloop(),
so no problem to maintain.

>> 14-Apr-99 14:12 Vlad Harchev wrote:
>> >   But the way it's implemented now should be considered incorrect. IMO, 
>> > user
>> > should understand that changing options have a penalty and not to do this
>> > very often. Also, IMO the use of caching proxies with SOURCE_CACHE:NONE 
>> > won't
>> > make things messy/slow. IMO, we should implement the correct behaviour.
>>
>> Using of locally installed cached proxis is a good solution
>> but they have a validation/no-cache/expired logic
>> thich may be contrary to our needs (say, dynamically generated pages
>> are always set to expired).

> I assume the SOURCE_CACHE has its own kind-of validation/no-cache/expired
> logic, just not date-based.  But I admit I still haven't looked at the
> code.

Currently no validation et all. Just reloading for \,[,*, etc. at the moment.
But the implementation works! (we are fixing minore bugs in mainloop now).

>   Klaus