I think Dave's comments are worth a lot :-). I understand that both IDs and their mappings to resolvers should be technology-neutral. I would greatly welcome this.
My own position is based on a much more limited perspective: Does the identifier system plays well with standard software, or is it creating a walled garden for those installing custom LSID software?
http://wiki.tdwg.org/twiki/bin/view/GUID/LsidHttpProxyUsageRecommendation is a valid attempt to support semantic reasoners. It is, as stated in its purpose, for those desiring to use LSIDs at all, and thus by necessity asymmetrical. It is just a bit complicated (and perhaps error prone) to publish all your data under LSIDs, but then never use them when referring to your own data.
With a technology independent ID, you could use http consistently, without preventing LSID users from translating URL-based-ID to neutral-ID to LSID-based-ID.
Gregor