CGI putting stack in use limitations
katir at hindu.org
Tue Mar 9 23:43:25 CST 2004
First to Jacqueline, Rodney and Monte and every one else contributing
in this area.
Question: looking through the "Intro to Rev CGI's" presentation notes
that you did, Jacqueline, we don't see any "disclaimer" on the the
possible issues arising from opening stacks with huge amounts of data.
(maybe i missed it) libCGI is touted as "able to handle any number of
stack and CGI's."
I presume, correct me if I am wrong, that, say if we installed rev on a
managed dedicated server on some host's farm with only 256K RAM (yes,
people are still running and trying to sell machines with that little
memory!) and we were to
start using stack "tooMuchData.rev" ## being a stack that was 500 K in
size, which is nothing for this G4 titanium...
the server would attempt to load the full 500K of stack data into
memory and that, we would sink the ship... am I wrong? What exactly
I suspect the answer is "Well of course, if you do something stupid
I guess the question behind the question is bigger: just how far can we
take this... using only xTalk and not bothering with mySQL or
PostGreSQL, forget LAMP... before we see performance problems that only
a macho relational database can handle. but it hardly seems worth it to
get my head wrapped around PostGreSQL when we're just try to serve up
500-1000 K of text data. That's not a lot.
For example, a real world application that would interest us: create an
on-line lexicon of all the words in all the books we have published by
Himalayan Academy Publications with definitions, references "see also"
links etc. I have few options: either build 26 html pages, one for each
letter of the alphabet... then, use HTDIG to index those pages and
.... well, it's just too painful and unmaintainable... delegate this
tedious work to some poor soul to work on for a week or two... if the
book editors say "oh gee, we updated these words..." then suddenly 20
man hours of work becomes 21 hours of work... on the other hand doing
an XML dump of each books glossary from InDesign and parsing that into
a stack is as easy as sweet lassie (water, yogurt, honey, touch of salt
and lime juice, two cubes of ice ;-) put the stack on the server and
then what... start using it and sink the ship.? OK. so, scratch my
head. output CVS and upload to a mySQL dbase and build a rev front end
and tell users to download a desktop app and forget their browser...
that also has problems... mostly the download special app resistance
factor, aside from using a macho dbase for such a limited record set.
The cgi access to stack data with a dynamic page build response via
browser to queries, seems so "cool" maintainable, editable, for small
data sets, but the challenge of loading the whole stack into memory is
Himalayan Academy Publications
at Kauai's Hindu Monastery
katir at hindu.org
More information about the use-livecode