Re: HURDNG : Which type of OS design could we have to thinknowadays ?

[Nigel Williams]

On Mon, 07 Aug 2006 19:46:48 -0400, "Jonathan S. Shapiro"
<address@hidden> wrote:
>On Mon, 2006-08-07 at 10:43 -0600, Nigel Williams wrote:
>> With hurdNg and coyotos I would gladly see legacy support de-valued (even 
>> jettisoned).
>
>> A goal to support ... PHP.
>
>You do realize that these two statements are contradictory?

Not my intention, my original sentence was perhaps unclear due to lack
of emphasis and elaboration:

"...is for a scalable web-server implementation which offered an
application environment _comparable_ (_preferably better_) to ASP or
PHP."

In other words once a web-server foundation with the desired
characteristics (scalable, secure, robust, configurable, tunable,
verified, extensible and so on) had been constructed, then in order to
motivate wider-interest it would need to provide an environment
capable of supporting something akin to the current style of
web-applications, whether they be the current REST model or AJAX-style
(should it become popular). The continuation-based model of Seaside
(based on Smalltalk) is another approach worth consideration too.

The exposed functionality need only be "comparable", at least in the
sense that the development model that was surfaced was broadly
familiar (to ease transition) to the legions of current web
developers. "Preferably better" was intended to posit a web-server not
shackled to the current model of serving files out of a hierarchical
directory system, which hyperlinking, inheritance, or file inclusion
attempts to re-cast (poorly) into a more relational-style navigational
structure.

Current web-server technology requires considerable expertise and
resources to reach even a modest level of security. Providing a
web-server that was secure out-of-the-box and which remained secure
without constant patching would attract considerable interest.

>PHP, Perl, Python, and friends require a depressingly large degree of
>legacy support. They all want something POSIX-like. Actually, it's not
>so much POSIX per se as certain *ideas* that are common to existing
>systems.

For sure, and given that these examples do not currently adhere (or in
some cases could not adhere) to a capability security model built on a
orthogonally persistent system I expect they would be left behind
closeted in a virtualized sandbox. 

>
>For example: try to imagine implementing any of these languages in a
>system that does not have a conventional file system...

I agree quite intriguing. It is interesting how many current
web-server exploits rely on poor ".." directory navigation semantics
which would never exist in such a system.

The outlined scenario is proposed as a short-term deliverable, within
n years, where n < 10. Ultimately I would hope delivering content over
the internet could be supplanted by something more sophisticated and
engaging like Croquet or its successors.

nigwil