Having done some thinking, I believe the best direction forward is now
to focus our energies on Sam and Dave's "simple modules" proposal,
making sure it meets all use cases including those of a secure subset.
Last week, I wrote up a candidate strawman wherein module singletons
would be layered on top of purely generative modules:
What I found is that, due to the different requirements of how modules
are named and what these names designate, modules written to one
pattern are not usefully modeled as a special case of the other
pattern; they are really just two different ways of thinking about
modular code. The result is that singleton modules with support for a
secure-mode loader that only accepts transitively immutable modules
(as originally suggested by Sam) is just as good for secure code as a
pure maker-style implementation.
The work is not complete, by any means, but I think this realization
means we may make quite a bit of progress.
Hi folks,
Having done some thinking, I believe the best direction forward is now
to focus our energies on Sam and Dave's "simple modules" proposal,
making sure it meets all use cases including those of a secure subset.
Last week, I wrote up a candidate strawman wherein module singletons
would be layered on top of purely generative modules:
http://wiki.ecmascript.org/doku.php?id=strawman:modules_harmonic
What I found is that, due to the different requirements of how modules
are named and what these names designate, modules written to one
pattern are not usefully modeled as a special case of the other
pattern; they are really just two different ways of thinking about
modular code. The result is that singleton modules with support for a
secure-mode loader that only accepts transitively immutable modules
(as originally suggested by Sam) is just as good for secure code as a
pure maker-style implementation.
The work is not complete, by any means, but I think this realization
means we may make quite a bit of progress. Cheers,
Ihab
--
Ihab A.B. Awad, Palo Alto, CA
Having done some thinking, I believe the best direction forward is now to focus our energies on Sam and Dave's "simple modules" proposal, making sure it meets all use cases including those of a secure subset.
Last week, I wrote up a candidate strawman wherein module singletons would be layered on top of purely generative modules:
strawman:modules_harmonic
What I found is that, due to the different requirements of how modules are named and what these names designate, modules written to one pattern are not usefully modeled as a special case of the other pattern; they are really just two different ways of thinking about modular code. The result is that singleton modules with support for a secure-mode loader that only accepts transitively immutable modules (as originally suggested by Sam) is just as good for secure code as a pure maker-style implementation.
The work is not complete, by any means, but I think this realization means we may make quite a bit of progress.