Dmitry Turin wrote:
> Julian,
> 
>>> RS> there is some fundamental flaw that makes a "fast" implementation impossible
>>> What flaw ?
> JR> I think the point was that you won't know about a flaw like that until
> JR> you have tried to implement it, thus it would be unwise to standardize
> JR> it earlier.
> (1) Theorem of existance of _fundamental_ flaw is not proved.
> (2) You postulate, that _fundamental_ flaw must be, even not naming (entitling) them.

I have no idea what you're talking about.

> JR> it would be unwise to standardize it earlier.
> Agruments are necessary, as usual ! Without them, it's only words.
> 
> But argument against are the following:
> large manufactorers (e.g. Oracle, etc) refuse implement before standardization.

On the other hand, depending on the standards body you use, you may not 
be able to standardize without showing at least one reference 
implementation.

>>> RS> Perhaps one of the key concepts is covered by a patent held by a jerk.
> JR> Then the specification would be of less use, because fewer people would
> JR> be interested in or capable of implementing it.
> (1) Ideas (instead of algorithms) are not capable to be patented !
> (2) International patent does not exist
> (if you want to patent, you must make this in each country separetely).
> 
> From where you have this fear ("perhaps") ?

If there are patents that are valid, you may either have to pay license 
fees, or you won't be able to release that product in the country where 
the patent is hold. When you're doing open source, you may be even in 
bigger trouble because your open source license may be fundamentally 
incompatible with the license terms of the patent holder.

I just wanted to explain why people care, that's all.

Best regards, Julian