Hi Pedro, > B). If the user has already downloaded the UMLS isn't that already indicative > that they had a valid account? As I understand it (I wasn't around at the time) this per-user licensing with a jit check was the deal that was worked out with the NLM. I think that repackaging and redistributing any form of the UMLS was not (legally) done before ctakes worked out the current arrangement. I think have heard ytex had an initial check upon installation, and we have talked about (would like to) use this model. The only drawback is a "single download, multiple install" site distribution possibility - which NLM didn't like. My information could be woefully outdated or just plain wrong. If anybody out there knows better then please chip in.
Sean P.S. If anybody would like to try to advocate a different arrangement with the NLM then that would be great. -----Original Message----- From: Pedro [mailto:teixeir...@gmail.com] Sent: Thursday, May 14, 2015 9:43 AM To: dev@ctakes.apache.org Subject: Re: UMLS Authentication failing despite correct username and password Agreed. Doing a direct string comparison seems like it will just break at the very next update. A). A check to parse the XML result looking for a result tag and that the contents are "True" seems better B). I'm not familiar with the history of that particular check but it seems overly restrictive to require a valid UMLS account check for every single run. If the user has already downloaded the UMLS isn't that already indicative that they had a valid account? I realize there are more ways around it in that case but requiring an internet connection just to run one of the UMLS analysis engines seems... suboptimal. Thanks for all the help sorting this out!