Yann Stettler wrote:
>Interesting claim. Sorry but I can't accept it myself...
>The resources/time needed to transfer such little amount of
>data via a pipe between two program is simply not relevant !
I'm not sure why you would focus on the time to pipe the
data. I don't remember ever making that the issue.
>Actualy, doing a good front-end could let you gain a lot of
>resource/time : Cache the result of the search so that
>when the user request the next few results, they can be
>returned immediatly without Swish-e having to search the
>whole index again.
The search time is not the constraint. Swish-e is pretty
fast; otherwise it would be pointless to bother doing
the entire search every time a page is displayed.
Caching is an option, as long as the storage options
are faster than a plain search. Let's say you store the
results in a text file. How do you track it in the
next "hit"? Cookies, maybe? URL-Encoding? When do you
clean it up? HTTP is stateless, you know. Assuming you
solve these problems, how "immediate" is it to
"set your cursor"? The swish-e results are not
standard in length. That means that you can't just
seek to byte position X to get to the listing Y.
You have to read each line into your script, literally
throwing away the junk until you get to the meat.
No problem if it's a 15 line file. What if it's a
2000 line file and all you want is the last 10
lines? What if you have 100 simultaneous connections
all doing this at the same time? What if the client's
cookies are disabled and you don't want to have someone
altering parameters in URL's? Caching in a
database isn't any better, and presents even
more possible cleanup problems.
You don't have to believe me. I've tried it
several different ways and I've found that having
swish-e limit its searches gives me the fastest
results. YMMV. It's pointless to argue religion, so
I'm ending my participation in this thread now.
You don't seem to be upset that swish-e can limit
the number of results returned. It seems pretty
silly to be arguing so hard against setting a
starting point as well, not to mention constantly
implying that anyone who perceives a need for
this functionality must be a poor coder. I've
given this patch out to enough people to know
that there's a demand for the functionality,
which is what Roy was asking about in the first
Received on Fri Jan 22 16:34:03 1999