[cap-talk] How desirable / feasible is a persistent OCAP language?

Sandro Magi naasking at higherlogics.com
Fri Jul 18 18:10:07 CDT 2008


Stiegler, Marc D wrote:
 > On a normal day this is really cool. But on a bad day -- and bad days
 > occur too often to just write them off as anomalies -- this can be a
 > serious problem.

While I agree upgrade is a tough problem, I think it's one that a 
compiler or runtime should have enough semantic information to solve (or 
it should not be too much of a burden to specify sufficient information 
to upgrade).

I have not yet studied the literature in depth, but researchers have 
built many dynamically patchable systems in C and ML, and upgrading 
persisted objects should be a straight extension of live upgrade. If it 
can be done for C, I can imagine it could only get easier for a language 
with more type information.

Here are the references to these systems that I've saved for later study:

http://www.cs.umd.edu/projects/PL/dsu/
http://citeseer.ist.psu.edu/duggan01typebased.html
http://www.lfcs.inf.ed.ac.uk/reports/01/ECS-LFCS-01-425/
http://www.cs.umd.edu/projects/dsu/DSU-TR.pdf
http://lambda-the-ultimate.org/node/2811#comment-41760

Sandro


More information about the cap-talk mailing list