On Thu, Dec 9, 2010 at 12:16 PM, Sven Van Caekenberghe <s...@beta9.be>wrote:
> > On 09 Dec 2010, at 20:39, Eliot Miranda wrote: > > > On Thu, Dec 9, 2010 at 11:14 AM, stephane ducasse < > stephane.duca...@free.fr> wrote: > > Hi guys > > > > When I see that. I would really like to see how we can create something > around the cloud. We have a lot of assets. > > Does anybody have some interest bootstrapping a business around that? > > > > _very_ I'd love to do high-quality VM work funded in this kind of > context. > > Isn't / shouldn't Cog already (be) a high quality VM then ? > Its code generator is naive, the object representation is over-complex, the GC is poor (no incremental mark-sweep), the FFI is weak (no threading), there is no multi-core solution available as yet, there is no 64-bit solution, there is no adaptive optimizer. While I'm making progress on some of these I certainly don't have enough time to focus on them and would love the chance to do just that, working with people like Igor and Marcus. > I don't think (I am actually pretty sure) that Ruby has a good VM, > Smalltalk's is way better (many more decades of experience, from the times > when hardware mattered). Up until very recently (and even after the RoR hype > started) they had no unicode support and no multithreading. Even now, they > always (have to) deploy clusters of multiple VMs to scale. [On the other > hand, this makes every deploy better scaleable because they have to do it > from day one]. > > Seaside can do all that too (the scaling/clustering), it is sometimes > little things that cause problems. For example, many people have tried but > very little have succeeding in building a VM themselves that is as fast as > the 'official' ones. Headless support is another one. Better sockets and > streaming would also help. > > Cloud support is actually lot's of Unix/Linux level work combined with > automation. Smalltalk, Ruby, Lisp, Python, Java, .Net or even C, it doesn't > matter much. > I fear you're right. But I'm told that being able to scale is key, and that is in part a VM performance and tuning issue, not just a deployment configuration issue. > > But don't get me wrong, I do think Smalltalk has an interesting future. It > is still one of the nicest languages/environments/IDE's/communities to work > in. And Pharo has been a great step in modernizing Smalltalk. And Cog just > made everything incredibly much faster. > Somewhat faster :) There's potentially much more to come. best Eliot > > Sven > > > > >