I'm considering using v8 as the main control language in a server
application (embedded in a relatively simple driver layer). However,
in this context the limitation of only using it from a single thread
in the process is putting me off, so I'm wondering how much work would
be involved in creating a more thread-safe version.  I'm happy to
accept limited communication/data sharing between v8 contexts running
in different threads - I think the application should be in charge of
providing any required synchronization there, rather than including
locking code in v8 itself.

Ideally, I guess this would be done by making all the code fully re-
entrant (with a layer on top to match the current interface for those
who don't need to run more than one context per process), but I
imagine it might also be possible to hack something together using an
OS-specific thread-local storage system.

I have four questions really:
1- How much work would be involved in this?
2- If I were to attempt it, where would it be best to start?
3- Are there likely to be hidden gotchas in the way generated machine
code works (or anywhere else I suppose)?
4- Is this something that other people would find useful?

That's probably a lot to ask, but any thoughts would be appreciated.

Thanks,

John B
--~--~---------~--~----~------------~-------~--~----~
v8-users mailing list
[email protected]
http://groups.google.com/group/v8-users
-~----------~----~----~----~------~----~------~--~---

Reply via email to