Re: ARC patent - Mailing list pgsql-hackers

On Mon, 2005-01-17 at 14:02 -0800, Joshua D. Drake wrote:
> >IBM can NEVER sue customers for using infringing
> >code before first informing them of infringement and
> >giving reasonable time to upgrade to uninfringing
> >version.
> >
> I can see it now:
> 
> We won't sue you (customer) but you have to upgrade
> to DB2 ;)

This is panic and is wrong-headed. They haven't even sent a letter
yet...

If we believe in this project, then ultimately, we should be aware that
the future *is* litigation, just like with Linux. Successful
people/projects/companies will at some point have to play hardball.
That's nothing to run scared of, unless you feel you have or will do
some harm to another.

Tom's view seems correct. IBM have *applied* for a patent; the community
is now aware of this and must plan accordingly. I see no reason to
contact IBM; they have no basis to complain as yet. If they had wished
to protect their patent they could have done so earlier - the dev
process here is open and visible, so there is a reasonable onus on them
to perform some form of minimum attentiveness on us if they see us as
competition. I have no reason to believe they do and our current
understanding is that IBM supports Open Source and therefore this
project. We support AIX, Linux on PowerPC, Linux on S/390, jdbc on WAS
to name but a few things IBM would be very happy with.

The patent has not yet been granted and seems to have been pending for
at least 18 months. We therefore have reason to believe there is some
chance it may not be granted, related prior art on buffer management
stretching back more than 30 years. By taking reasonable actions now we
will buy ourselves reasonable time should it ever be granted.

It seems clear that anybody on 8.0.0ARC after the patent had been
granted could potentially be liable to pay damages. At best, the
community would need to do a "product recall" to ensure patents were not
infringed.

So, it also seems clear that 8.0.x should eventually have a straight
upgrade path to a replacement, assuming the patent is granted. 

We should therefore plan to:
1. improve/replace ARC for 8.1
2. backport any replacement directly onto 8.0STABLE as soon as any
patent is granted

Point 1 was going to happen anyway, so there is really less to worry
about. ARC is a better idea; it is likely there are even better ones.
ARC says nothing of how to clean the LRUs of dirty pages, nor does it
specify how to scale the algorithm to multiple CPUs.

The code already supports such a migration from 8.0.0 to 8.0.x

If any community members are planning selling products derived from
PostgreSQL 8.0.0 then it might be in your interest to put some money in
the pot for a legal fund and also to fund dev of a new buffer management
strategy. If those community members wish to delay release of their own
derived products then that's up to them.

-- 
Best Regards, Simon Riggs