Read with interest Tanmoy and your comments on e-journals. I think he misses a crucial point in the discussion regarding who archives what etc. But then so do you in your remarks about tokens. It seems to me that the key point is that:
A journal is a selected collection of links.These links are to be sifted out of a number of submissions by a bunch of referees and editors. It is up to the author to ensure the readability and portability of these links.
As regards volatility--perhaps the real solution is to let the original documents stay on the disks of the writers. As regards the possibility of tampering with a paper the magic-cookie idea could be used here to authenticate a particular version of the paper.
An author could then also re-locate his link by sending a message to the corresponding journal which could then check (via magic-cookie) that this re-located paper is indeed the original one. (Is there a possibility of having some redundancy built into the HTML stuff so that if one can't reach one node then another can be made available?)
The advantage of this scheme would be that:
The main problem as yet un-addressed is that the idea here is being mooted by a bunch of young(?)sters. How does one get a sufficient number of accepted seniors to agree to be referees? How does one ensure that such "publication" will be given sufficient value in the schemes of evaluation? Perhaps the answer is that this could be run as a duplicate system (in addition to real (rain forest) paper publication) until it acquires sufficient strength--this can only work if the electronic journals do not insist on "copyright".