[oc] Voting [WAS: Re: Inquiry]

[Marko Mlinar <markom@opencores.org>]

> What might be a better way to measure the quality of an IP core,
> is to list the number of real implementations it has seen. This
> type of information should be easily obtained from real users and
> verifiable as well. For example the Voxi success story is a good
> reference for a hand full of IPs from OC. I believe the OR1K also
> has seen a few other implementations as well.
>
> > Eventually cores that are more popular and are of better quality will
> > gain more votes than others.

I would not vote for a core that is useless. Note that this way you still 
cannot cover e.g. 10% of all votes, which would some cores surely get if the 
users would not be too lazy.

> I don't think popularity will necessarily reflect quality. I think
> a core like the UART will be more popular than the AC97 controller,
> just because there are more applications and users of an UART than
> of the AC97. This does not say anything about the quality of either
> core. However, a chart of the "popularity" would show the UART
> somewhere on the top and the AC97 on the bottom.
>
> > Perhaps this can help giving an impression that cores are really useful
> > and used in projects every day.
> >
> > Marko
>
> Quite honestly I think it will be tough to find a good solution
> as any type of judging will be objective and we really want to
> make it based on facts and not feelings !

Actually I was aware of the above fact, but maybe if we define different 
metrics, one can estimate what is happening. For example in uart case you 
would get:
(just for example, nothing to do with existing cores on the net)

the idea is good: 100
I have downloaded it: 300
core of a good quality: 200
I have tried it: 30
I have tried it and it works: 25

and in case of AC97:
the idea is good: 10
I have downloaded it: 30
core of a good quality: 25
I have tried it: 15
I have tried it and it works: 12

I think we can clearly see the quality and popularity of the core. Of course 
having 5 metrics is too much, but maybe we can define just 2 or 3 to get good 
results.

Furthermore, the above voting would not be to get a top 10 list of the cores 
or the competition, but to rather differentiate finished and good quality 
cores from projects with the code that does not even compile...
So when displaying them on the web one one can rank them according to the:
"I have tried it and it works" divided by "I have tried it"
or just list them alphabetically above certain threshold.

Marko


--
To unsubscribe from cores mailing list please visit http://www.opencores.org/mailinglists.shtml