From: Tom O. <tm...@eb...> - 2004-06-18 19:46:04
|
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Benjamin H Szekely wrote: | | Hi Tom, | We did initially have that in our registration. We have some | reasons why we limited registration to single LSIDs. | | 1.) While it may be true that you plan on annotating several LSIDs | within a namespace, you will still probably annotate | a few at a time so the call wouldn't be that expensive. You misunderstand - you are making the fundamental assumption that all annotation is manual, this is incorrect. My metadata service could act as a function over the space of all sequences, and therefore have metadata for any current or future sequence with appropriate concrete data. There is absolutely nothing in the LSID specification nor philosphy that dictates metadata should be fixed for any given LSID, lazy evaluation is not only possible but desirable in many cases, by implementing a costly algorithm as a metadata authority we can minimize the computational load over time (cache on demand) and provide transparent access to otherwise inconvenient tools. Another example would be a functional prediction algorithm such as InterPro Scan - consumes a protein sequence and emits a set of cross references to suggested functions within an ontology. The scan itself is computationally expensive, if we were to provide this functionality as a metadata server we can avoid ever having to perform the scan until someone explicitly states an interest in the data, subsequent calls would use the cached information. By allowing these kinds of metadata the vision of LSID being used to build navigable webs of knowledge is a possibility - precluding this functionality would in my view be a major mistake. Cheers, Tom Oinn -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.7 (Cygwin32) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFA0x3/vIo3NIa9d1ARAlndAJkBEpu86OFAO7guFC1XqC+WBgXIbwCgnX+D rcMYUU6WK9Any0l62gyZOjY= =p0BC -----END PGP SIGNATURE----- |