NCBI gets rid of the old GI identifiers for sequence records by September 2016, see here.
would all existing entries change or only the new submissions?
Appears to be all entries. GI's may still be used internally at NCBI but they will not be visible/included in GenBank, GenPept, and FASTA records.
Indeed, as far as I understood is going to be eliminated from all entries.
Although they now provide accession2taxid map files, these do not work with makeblastdb -taxid_map option even if you parse columns 2 and 3 (accession.version & taxid) into their own file. It's relatively painless to join the taxids to tabular blast output, but it gets less convenient if you also want to see e.g. "sscinames" ;-( There should really be taxid2alltaxinfo map files..
I suggest that you send that request in to BLAST help email address. Perhaps NCBI will start making that file available by the time GI gets phased out in September 2016.
I did that yesterday. At least right now there is no solution and you have to do post processing yourself.
This. It's crazy how many slow API alls are required to extract meaningful taxonomic information from BLAST results. It's hilariously user unfriendly. Thanks for making the request.
So it looks like we'll soon have to do this to get any meangful info from BLAST results:
Someone please tell me there's a better solution