> No.  You can't review articles.

What???? We can, and do, all the time.  You may really mean we cannot
(without some cost) administer this process.

> You can't provide easy searching and access
> to all articles in an area.

This is certainly something that should be provided a t a reasonable
cost by the correct type of organisation.

But in pure maths this has only ever been seriously attempted by Math
Reviews and Zentralblatt; and both of these, although naturally having
close connections with publishing, are clearly part of the mathematical
community and employi at senior levels academic mathematicians rather
than "publishers".

I have no idea if this yet another way in which maths is "different"
but that concept is not at all new to me; it is a daily part of my
working life as I represent the species in an institution which
necessarily (and very effectively) standardises anything that moves
(or at least anything that costs money).

I do know that maths is very different in the requirements it makes of
intra-document search engines.  This is one, of many, reasons why the
idea of MathML came to fruition (NOTE: I am not claiming that MathML,
as is, has delivered what is needed in this area).  And again, here,
it seems unlikely that, even given the right languages/tools,
publishers will be able to provide useful added-value in this area
without using specialised mathematicians to encode documents.