Username or EmailPassword
A ton of work has gone into this project. Seems destined for server work, but I had hoped that some of the system level changes would also be pertinent to the desktop... and that desktops similar to PC-BSD would pop up around DragonFly. So far it has not come to pass.
I think the idea of individual desktops running on top of some sort of dragonfly meta-cluster would be AWESOME.
I'm drooling just thinking about it.
but it would seem to me the server stuff needs to be done and stable before you could even start thinking about that.
Yes, I'm curious what part of that requires dragonfly?
I like the roadmap and the design of hammerfs, not sure what the real tangible benefits of them are exactly. When designing an entire system like this, its tough to keep up with all of the other performance improvements in oses with more manpower.
I think dragonfly needs to a better job of "selling the dream" with specific use cases and a plan of how to get there from here, to enlist some fellow dreamers. Maybe even some dreamers with a business plan that builds on the dream.
I fear Dragonfly may end up being Plan 9 V2, a technically superiorly designed os by brilliant engineers, that isn't better enough than its predecessor.
What is beautiful about DragonFly is that is a set of new features or new approaches coming in an already usable system. They do a lot of very impressive things but their system is still running!
Their vkernel implementation is simply amazing, same than its HammerFS (ported to Mac OS X some days ago).
Yes, I would like more explanation about their differing approach to scalability and the potential benefits in a multi core environment over FreeBsd and Linux.
Something like this [forgive the technical mistakes I'm venturing into an area where i am not an expert]:
One of the features they mention is multiple processes working on a single file in a non blocking simultaneous manner.
One potential application of it would be something like Mysql's innodb table space. If it were previously blocked form simultaneous writes to the differing tables within, allowing multiple simultaneous writes to it could potentially speed things along.
Of course because the changes are at the os level, they couldn't be implemented in a portable manner, but lets say their was a special version just for dragonfly. Is there a driver for enterprise level flash PCI cards like io fusion? No .. well then the storage system on DragonFly is at a disadvantage and the technically inferior Linux/FreeBSD sequential write methodology could still beat the crap out of it due to the faster storage.
I'm one of the (least active) DFly's contributors and I am following its development very closely.
Ok, so we will not conquer the world. Duh, who cares ? In the meantime, we are learning tons of interesting stuff, we are exploring new ideas and _we are having fun_.
The team consists of skilled, easy-going and very cool people. As for Matthew Dillon in specific, he is a coding god, and by saying this I severely underestimate his skills. Yet, he is insanely patient at explaining things, answering tech/usability questions, fixing regressions, etc.