Re: [TCLCORE] Upgrade of Tcl's String Hash Algorithm

[Eric M. <er...@el...>]

Donal K. Fellows wrote:

> Hi everyone
> 
> This is just a short note to let everyone know that I've just upgraded
> Tcl 8.6's string (and Tcl_Obj) hash algorithm.

I'm not much involved in the development of Tcl anymore, but shouldn't a 
change of this magnitude require a TIP?  The consequences of changing 
this algorithm are significant, yet you present this work as a fait 
accompli, with the implicit assumption that it is simply understood that 
the existing Tcl hash is bad and in need of replacement.  Do you have 
data to support that conclusion?  What are the specific scenarios in 
which the existing hash is inadequate?

> How much better is the new function? I'm sorry to say that the state of
> research in this area is remarkably poor; very few people are looking at
> what makes for a good *non-cryptographic* string hashing function.

This statement does not give me any confidence either in your analysis 
of the deficiencies of the existing hash or the advantages of the new 
hash.  At the very least, it seems you should be able to demonstrate the 
superiority of FNV over the existing hash for the specific problem 
scenarios that prompted you to change the hash in the first place.

Bob Jenkins has spent a lot of time thinking about and designing hash 
algorithms.  Did you look at his web page here:

	http://burtleburtle.net/bob/hash/index.html

Or his article in Dr. Dobbs Journal, reprinted and expanded here:

	http://burtleburtle.net/bob/hash/doobs.html

There are some (albeit brief) comments on FNV in the latter.

Best regards,

Eric Melski