Saturday, November 17, 2007

IDTP: too early or not too late...

We had a very useful discussion yesterday with Teravus on IRC regarding my IDTP experiments.

I'll try to squeeze the guts into a RegionModule - looks like it is a very nice place, and could be made loadable by those who want it.

There's however a very interesting "philosophical" point which we discussed. The point of view that it is too early for such a thing - given that a lot of "basic" stuff does not work yet. This I totally agree with (that the basic stuff does not work), however, the areas of trust and interdomain relationships will require a lot of further thinking and experimenting - I would be naive to tell that the IDTP itself would solve the world hunger. No, IDTP merely opens up a large can of worms, and it would be a huge pain for the users if it was introduced later as an additional "feature" once the things do work. And given that everything would be hopefully solidified by that time, any incompatibilities will result in "chewing gum and matches" workarounds - which is what plagues so many RL software items - "get the features done first, and the security thing, we'll take care of it once we get the functionality done".

For my own memory, the worms in a particular order:

- the trust model - first of all, how do you ensure that John Doe from NowhereSim is really John Doe ? Do you trust that NowhereSim is NowhereSim, and not a rogue impersonation of that ? Do you trust that the sim SweetHome, to which you are going to TP, is really the SweetHome, and not EvilJerks, pretending to be SweetHome ? What do we do to avoid the need of all-to-all sim authentication (well, social network of trust seems to be the most appropriate approach to me here)

- policy enforcement glue for the grid mode - even though there the trust is largely assumed anyway - which is why in my opinion the grid mode is not workable - all it takes is a single miscreant, and the whole concept will be poisoned forever.

- the "little" detail of distributed inventory - right now the assumption is that there is a central repository, which simply pulls takes the LLUUID, and spews out the contents, or takes the contents and LLUUID and stores them somewhere. This kind of abstraction is nice and convenient, but with the growing number of agents it quickly falls apart unless given a significant financial investment for the infrastructure - this is something that I'd consider as a potential risk - the existence of the central money-sucking infrastructure.point. I'd rather think the very same amount of money should be transacted by the entities comprising the "constellation", allowing each individual owner to establish their own policies (and prices, if necessary) for the resources. Then the market would take care of itself, and if there is The Best inventory hoster, or such - eventually they would be able to come afloat and have a sustainable business model.

- micropayments - well, I do not even go there yet. it's a meta-worm - with some annoying RL tentacles.

In a nutshell - I'll start refactoring the code as the RL permits (not very promising), and will try to stick it into a regionmodule.


CB said...

You have been

CB said...