[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: A CSP model for Java threads



Oyvind,

[Apologies to both lists for the off-topic material.
I'll try to make this my last reply before taking this off-list.]

> 99% of the developers here use networked PC's running Win95, Win98
> or NT. Installing Linux and rebooting to perform FDR checks isn't
> very attractive.

So you put a Unix box somewhere on the network and access it using X
or Vnc.[1] A suitable box costs a fraction of the licence price.

> Couldn't you have some kind of multilevel pricing? 

We do. There are evaluation licenses, short-term licenses, full
licenses and site licenses, with upgrades possible.

We've talked about some sort of `FDR-lite', but it is very hard to see
how to restrict it effectively; if we bound the problem size in any
way, then people start distorting their CSP to try and work round the
limit. We need more good CSP case-studies, so the last thing I want is
to give people an incentive to write bad CSP.

> I do hope that FDR will be pressed down in price by tools like
> LTSA, even if I absolutely don't want a free tool! But isnt N*8500 
> the same as  xN*(8500/x) ?

No. That ignores the cost of providing support.  One reason why the
academic licence is comparatively cheap is that we don't have to
handhold new CSP users; that's the job of the lecturers and
demonstrators. Unbundling support for the commercial version didn't
work.

   Bryan


[1] "The network is the computer" ?