The thread is now here..
Well, any more ideas?
I mailed kooma a couple of times but he hasn't replied yet.
---
have to get to 30... have to get to 30... ;)
Printable View
The thread is now here..
Well, any more ideas?
I mailed kooma a couple of times but he hasn't replied yet.
---
have to get to 30... have to get to 30... ;)
give up no 30 for you...
Kooma had said he was going to be gone for a couple days... What ideas did you send him?
I was just passing back a few ideas about the internal database structure.
kooma said he was thinking that people would submit the entries in HTML and then put the keywords inside of meta tags... which sounds pretty reasonable to me. But we were talking about different things, I think, by accident :)
I was thinking we could either be really cheap and have the database look like this on the inside:
keyword, keyword, keyword, etc..
link
%
or we could do something like this (probably a second version after the first):
<entry printf>
<key>printf</key>
<key>vprintf</key>
<article>cdb/c/printf.html</article>
<related>sprintf</related>
<related>fprintf</related>
</entry>
which would be much harder but more flexible. Do we need this kind of flexibility? I think it would be nice, but maybe something for after running on the first type for a while.
Another bit is that because everytime you make a cgi request the search program is run anew.. so it has to reload the database and search through each time.
A possible way of making it faster would be to have a server program run all the time waiting for connections on a port from the search program, and this server would have the database in memory at all times.
What do you think?
Can't help with the search, but I agree about the flexibility issue. If the first way is easier, we should leave that up while we actually do work on the final version.