Anyone have any experience with SkyOS?

[kamper]

Originally posted by: VirtualLarry

Originally posted by: kamper
What's so hard about generating a unique id? Have you seen the uuid spec?

Actually, no. Link? The problem is not necessarily just generating a unique one, but a tamper-proof, provably unique one.

Here's some light reading for you
There're probably shorter documents around. Long story short, it's an algorithm specifically designed to produce numbers that are "guaranteed" to be unique.

I believe it takes into account your machine's MAC address, the date and time and, of course, a random component. If you generate a number correctly you are virtually assured that no one on earth will generate that same number, so long as they also follow the rules.

I don't know a whole lot about the algorithm but it actually concerns me that it's only 128 bits. Sure that's big, but IPv6 addresses are 128 bits too. Anyways, I suppose it would at least be reliable for the purposes you describe. Where I work, we use uuids as keys for our database. Instead of letting the database generate an autoincrement, we generate the key in the app and insert it, assuming that it will be unique (hasn't failed yet ).

 

[robisc]

Wow, this thread did take a turn for the worse, as for "alt" Oses, I have paid for and tried the latest SkyOS betas, it is OK to play around with now that thay have the TCP/IP stuff worked out, but is is far from being a complete usable OS IMO. As for Yellowtab, I am a BeOS fan and had high hopes for this OS to replace BeOS R5, but I haven't been able to get the latest beta to install on the couple boxes I tried it on, it does seem that Yellowtab ver 1. will be released at CeBit 2005 coming up next week, maybe it will work.

 

[Netopia]

Hey robisc, where can one get a copy of the Yellow Tab beta?

Joe

 

[robisc]

Hey robisc, where can one get a copy of the Yellow Tab beta?

I bought it early on from their website.

 

[Sunner]

Originally posted by: Netopia
I have an ongoing love affair with BeOS... at least with the whole history and concept of it is/was just fantastic and WAY ahead of its time. Ahh... what could have been but isn't.

Joe

Yeah, it was a very very nice OS, once you got it working.

 

[Netopia]

Ok, cool. Just didn't know if there were any open betas that could be d/l'ed and tried.

Joe

 

[VirtualLarry]

Originally posted by: kamper

Originally posted by: VirtualLarry

Originally posted by: kamper
What's so hard about generating a unique id? Have you seen the uuid spec?

Actually, no. Link? The problem is not necessarily just generating a unique one, but a tamper-proof, provably unique one.

Here's some light reading for you
I believe it takes into account your machine's MAC address, the date and time and, of course, a random component. If you generate a number correctly you are virtually assured that no one on earth will generate that same number, so long as they also follow the rules.

Thanks, I appreciate you taking the time to look up the link, but it looks much the same as MS's GUID scheme (I assume MS's is based on that spec), but that's not good enough for what I was describing, not by a long shot. That's only "unlikely to contain a duplicate id", not "guaranteed, provably unique". Big difference.

Originally posted by: kamper
I don't know a whole lot about the algorithm but it actually concerns me that it's only 128 bits. Sure that's big, but IPv6 addresses are 128 bits too. Anyways, I suppose it would at least be reliable for the purposes you describe. Where I work, we use uuids as keys for our database. Instead of letting the database generate an autoincrement, we generate the key in the app and insert it, assuming that it will be unique (hasn't failed yet ).

Yes, but now imagine that this "datebase", is the underlying object-store for the entire computing world, globally, worldwide. Imagine what would happen, if some of the objects and/or node entities within the system started to show up as duplicates. I would consider it much like a hanging thread from a sweater. If one should happen to pull it... well, bad things would happen to the system. I guess it really depends on what your needs are, if they are just "mostly unique", and what the worst-case scenario that could happen should any duplicates actually show up. (Keeping the system's usage of those ids, and thus damage from failure, localized, is one solution.)

 

[kamper]

Oh, I didn't know you were about to create a global file system and were just held back by this one detail I'm certainly not an expert on uuid's but I was under the impression that they had enough theoretical work behind them to make them quite reliable and certainly by the time that you want to build a very large distributed file system more research will have been done and larger uuid's will be used.