From my understanding, both tools blastdbcmd and blastdb_aliastool would allow the user to subset a database from GI numbers of a bunch of sequences using them as follows:

> blastdbcmd -db nr -entry_batch human.gi -target_only -out human.txt
> blastdb_aliastool -gilist human.gi -db nr -out nr_human

Apart from one of them giving as result a file with sequences and the other one an already formatted database, is there any other difference? Speed/performance, maybe?

Could the output of blast_aliastool command be used alone or would it need the rest of the nr database to be uncompressed for it to work? In the latest case, the first way of producing a subset of a database -although involving an extra step of making the database again from the sequences- will be preferably if the disk space is limited.