From what I remember here are a few:
1) Indexes will need to be versioned to track
a) module changes -- If a module changes, then the index will most likely be off. I say most likely because there are scenarios under which it will still be correct. I think the risk of having a bad index because it was not rebuilt is so high that the practice will be to rebuild when the module changes.
b) application changes -- Lucene gives a lot of flexibility as to how an index is built and what is searchable, retrievable or stored in it. Any change in that strategy probably will require a new index.
c) lucene library changes -- An index may be upwardly compatible, but not backwardly compatible. Because of this risk, I think it will be necessary to build the indexes with each version of the library that is currently being used. This probably will require a policy of the previous version, the current version and the next version, as used by Sword. (Or something like that)
2) The index built for sword is not usable by JSword and visa versa. At this point we are not sure why this is. It could be that the c and java versions of lucene don't build the same index for the same strategy or it may be that Sword and JSword use different strategies in building the index. In the earlier JSword threads, Sword was not considering having downloadable indexes. Since it is being discussed, perhaps we should collaborate on looking at and perhaps solving this issue.
3) The building of indexes is compute intensive. Building them on the crosswire server will need to be done in a manner so as to not disrupt the services that crosswire provides.
4) We cannot presume our users have significant computing power. At least we have users with old laptops who want to use JSword. They may be limited by CPU Mhz, RAM and disk.
Chris Little wrote:
Lynn Allan wrote:
... stuff deleted
I tried the "Create Index" for searching the ACV. It is still grinding and taking a loooooong time ... almost seems locked up, but I'll wait a while longer. Every minute or so, another block appears in the "Progress Indicator" so this could take a while. Seems a little more than half done after 13 minutes on an older 933 mhz PIII ... done @ 25 minutes. Is this expected behavior?
Yeah, for me, on a 1.5GHz P4-M, it took 20min to index the NASB. The ACV just took me 18min. If we can't get the time down to a reasonable level, we may post indexes for download (the NASB index is about 6.7mb or 2.5mb zipped).
_______________________________________________ sword-devel mailing list sword-devel@crosswire.org http://www.crosswire.org/mailman/listinfo/sword-devel
