Re: failure notice (fwd)

Oliver Xymoron (
Sun, 1 Mar 1998 09:44:13 -0600 (CST)

On Sat, 28 Feb 1998, Dean Gaudet wrote:

> Er well those programs do use more magic than just time() concat getpid(),
> but time() concat getpid() is the only part of the magic that is used to
> distinguish indentifiers on a single system. Generating universally
> unique tokens is a real challenge. Message-id collision is not that
> abnormal (which is one reason I don't trust it for duplicate message
> elimination, but that's off-topic).

A kernel API could be created to hand out unique tokens rather than 16-bit
PIDs. But a system that can do 10k forks/sec (Ultrasparc Linux) and can
have more than a gig of RAM probably needs support for more that 64k
processes anyway, no?

A library to generate a decent length universal unique ID from sources
like high resolution clocks, MAC addresses, and the random device might be
useful. With the number of networking and other applications that
currently need such things, a standard API to generate them might be in

 "Love the dolphins," she advised him. "Write by W.A.S.T.E.." 

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to