Re: [BioRDF] All about the LSID URI/URN

Phil

Well said!  As a practical user of LSIDs for all our infrastructure we 
are interested in building bioinformatics platforms that work.

We are preparing carefully developed argument about why we use LSIDs and 
why they have been proven invaluable. And what we would like to see done 
better.

And guess what - we don't look at LSIDs from browsers. We will post this 
tomorrow -- we would prefer a well developed discussion document be 
discussed on Monday.

This is an important topic because we really deliver middleware that 
bioinformaticians use to do biology - and its uses LSIDs. This is not a 
theoretical discussion for us.

Carole

Carole Goble
Director myGrid-Taverna project
Chair Open Middleware Infrastructure Institute-UK


>>>>>> "HST" == Henry S Thompson <ht@inf.ed.ac.uk> writes:
>>>>>>             
>
>   HST> Sean Martin writes:
>
>   HST> So, register one of lsids.org, lsids.net, lsids.name or
>   HST> lsids.info, and use e.g. http://lsids.or/xxx instead of
>   HST> URN:LSID:xxx.  Bingo -- no new tools required, works in all
>   HST> modern browsers :-).  
>
> But if the file you are referencing is, say 5Tb, then it doesn't work
> in a browser at all. With LSID's on the other hand, you may get back a
> choice of methods to access the data, including one which can cope
> with 5Tb of data. 
>
> Incidentally, the approach that you are suggesting demonstrates that
> LSIDs could be used in concert with URIs. In which case, putting
> "http://" into the LSID adds nothing. This is, in fact, exactly how
> DOI's work. But the resolution through the doi.org proxy is a
> convention which can be changed without changing the dois. 
>
> Sean is entirely correct that encoding the protocol for the transport
> layer and a DNS based resolution host into the identifiers is a recipe
> for instability; maybe not a problem for many things, but a disaster
> for many parts of bioinformatics. I do not still want to be doing
> synonym resolution when I am 60.
>
> What I have totally failed to understand about this discussion is why
> it has been couched in terms of whether we should use LSIDs or
> URIs. Of course, LSIDs non standard protocol in a pain, and the two
> step resolution adds latency. But, this is the cost for circumventing
> URIs protocol dependency which is also difficult. 
>
> We should just circumvent this entire discussion about which
> identifier is perfect for bioinformatics because I can tell you the
> answer straight out. None of them. LSIDs answer a need. So, people use
> them.
>
> Cheers
>
> Phil
>
>

Received on Tuesday, 25 July 2006 17:05:02 UTC