On Sun, 23 Mar 2008 14:08:17 -0700, Jeff Garland wrote:
To reiterate, its really easy to write a JSON archive that operates just like current archives, but the thing to determine is whether Boost Serialization can inter-operate nicely with the outside world. I think the answer is yes, with limitations. IMHO, any GSoC application should address this, but I am not reviewing them so don't listen to me ;-)
I'm reviewing them (as are others listening here) and really suggest they listen to you
LOL! Thanks for the compliment!
I was actually unaware of the object graph limitations in JSON. Of course it turns out there's at least one proposal to fix these problems:
http://www.jspon.org/?mode=html&noscript=true
Or the project could specify limitations to the types that can be serialized in the archive. That's up to the students to propose...
Oh, I'm liking the odds of this succeeding. I am pretty sure that a summer of 6 hrs/day is more than enough to implement a good chunk of a json_[io]archive and/or jspon_[io]archive depending on the availability of parsers. If the student was to concentrate solely on jspon, one of the tasks could be writing a working parser for another language (Java? Python?) to get the ball rolling... Just some pie-in-the-sky for Sunday. IMHO, this has the most promise of providing all features of Boost Serialization. While I have your attention, please don't make the library header-only. :-D -- Sohail Somani http://uint32t.blogspot.com