[cap-talk] Capability levels - transparent network extension, no encryption

[Jed at Webstart]

At 05:14 PM 8/17/2006, David Hopwood wrote:
>Jed at Webstart wrote:
> > That's why I published RFC 712 - essentially
> > an earlier version of the DCCS description (2/76).
>
>RFC 712 is one of the few RFCs that are not online. If you still have a
>copy (scan or transcript), I'm sure that <rfc-editor at ietf.org> would
>appreciate it.

Heh.  Somewhere in my basement I have a notebook 
that Jake Feinler sent out and that I kept 
populated with all the RFCs from the early days 
of the ARAPnet.  I remember it as the one with 
the red stripe on the top. Unfortunately I looked 
once already and didn't find this particular 
notebook, which I expect would have that RFC I 
wrote up in it.  I'll try taking another look 
(climbing into a crawl space in our basement) 
this evening if I get a chance.  I don't think 
there's anything relevant in it. That RFC appeared between:

J. E. Donnelley, "DCAS" - A Distributed 
Capability Access System, Lawrence Livermore 
Laboratory Report UCID-16903, August 1975.

and the later:

J. E. Donnelley, 
<http://www.webstart.com/jed/papers/DCCS/>A 
Distributed Capability Computing System, 
Proceedings of the Third International Conference 
on Computer Communication, August 1976, pp. 432-440.

that's on-line at:  http://www.webstart.com/jed/papers/DCCS/

and I believe the content is essentially the 
same.  There may have been some variation to get 
it into the form of an RFC, but from my memory 
there wasn't anything significant in that 
area.  Still, I would like to find it.  Perhaps 
at the same time I should look for that previous 
LLNL report from 8/75 and scan that in.  I'm not 
sure I still have a copy of that report, but I 
may have.  Also I expect LLNL has a copy and I 
could probably borrow one to copy - perhaps from 
a colleague still working there.

>[...]
> > What about the case, though, when the capability being sent "up"
> > to go out on the network is a local CCS capability that isn't a local
> > representation of a network capability?  In that case it would seem
> > that the network extension service must store the local CCS capability
> > for remote access and make up some sort of a YURL that it can
> > communicate through the existing YURL to allow the desired delegation
> > to happen.  Even beyond the data formatting issues (how do YURLs
> > show up in 'invocations' on other YURLs?) this business of turning
> > a CCS capability (e.g. consider CapDesk or E) into a network capability
> > of the YURL sort seems a bit problematic.
>
>I'm not sure why.  - in other words, why?

I'm afraid I can't give a quick answer to this 
question.  Perhaps I should withdraw the "Even 
beyond" for now while I think about it and focus 
just on the data formatting issues which seem 
formidable to me and seem to make extended local 
objects much less palitable for use at the higher 
levels than native designed objects at the higher level.

Regarding:

>There must be a mechanism to create a new object at the network level,
>assigning it a new capability representation (YURL, etc.) So just create
>a new network object with the behaviour of forwarding to the local object.
>This is often called a "scion" or "skeleton". Various optimizations are
>possible -- cacheing so that there is at most one scion per local object,
>shortening chains of scions and "stubs" (proxies), etc. A variant that
>also supports object migration and acyclic distributed GC is described
>in:
>
>   M. Shapiro, P. Dickman, and D. Plainfossé
>   SSP Chains: Robust, Distributed References Supporting Acyclic Garbage
>   Collection. Technical Report 1799, Institut National de la Recherche
>   en Informatique et Automatique, Rocquencourt, France, 1992.
>   <http://www-sor.inria.fr/publi/SSPC_rr1799.html>

I found the above paper interesting, particularly 
given the time frame and geography of its 
publication.  I wonder if somebody (perhaps 
MacrS?) could put this paper into context.  What 
motivated it (i.e. what object systems was it 
considered/designed for?) and what happened to 
the concepts (i.e. was anything implemented, and 
if so for what systems and what happened to the 
implementation?)?  I'd like to know a bit more 
about the context before diving into it more 
deeply as I found my initial foray challenging (terminology, etc.).

Let me just make sure we have a common 
understanding of what we're talking about here 
David - r.e. "extending" local objects to a wider 
context.  An example would be to consider a 
KeyKOS system on the Internet (was there 
ever?  An EROS or Coyotos system would do of 
course) and to make some of its capabilities 
available as widewords or YURLs or perhaps on an 
Amoeba network and made available as Amoeba 
object references.  I refer to the implementation 
and use of such an extension as "problematic" and 
you ask "why?"  Is that what we're talking 
about?  If not, then perhaps you could rephrase the question as you see it.

I don't know of any such implementation that's 
ever been done.  If there are any I'd be 
particularly interested to know if any such 
implementations have found practical application.

--Jed http://www.webstart.com/jed/ 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.eros-os.org/pipermail/cap-talk/attachments/20060818/990ebe4b/attachment.html