Elisa,
Thanks for your case description.
For my understanding, we can do it may be in following steps:
1)Get all the unique identifiers out and sort them alphabetically.
2)Save those information into Memory or disk file in a dictionary format(key is identifier, value is the recordId of the dbf).
3)When a identifier want to be queried , just do a binary search on those dictionary keys to get out recordId, this would be very fast with using binary searching algorithm.
4)Use GetFeatureById and get out what we want.
Of course, this is just a trick way, hope it works. Now, I and David are working together on the support of the dbf indexing, it is quite complicated, hope we can get it there in the coming weeks.
clicketyclick.dk/databases/xbase/format/ndx_example.html#NDX_STRUCT
Any more questions just feel free to let me know.
Thanks.
Yale