[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: fuzzy check sums
- To: Michael Grant <http://www.grant.org/~mgrant>
- Subject: Re: fuzzy check sums
- From: http://dummy.us.eu.org/robert (Robert)
- Date: Tue, 2 Apr 2002 20:00:40 -0500
- Folder: folders/o1.home
- In-Reply-To: <3http://www.grant.org/~CAA0B77.FF8A8FD7>
- References: <20020402171831.6248.qmail> <3http://www.grant.org/~CAA0B77.FF8A8FD7>
Sorry about the "spam" thing. I screwed up.
> From: Michael Grant <http://www.grant.org/~mgrant>
> Date: Tue, 02 Apr 2002 21:50:15 +0200
>
> I didn't post the code, nobody ever got back to me. What do you think,
> shall I post it to the list or what? Having read the list the last few
> days, it's more of a questions list and not a dev list so I wasn't sure
> if it was the correct venu. Are you one of the developers of dcc?
No, I'm not a DCC developer. I was mainly interested in your algorithm
since I have my own fuzzy checksum program. (See
http://dummy.us.eu.org/robert/software/fuzzysum/index.shtml .)
> I've attached 2 files, essence4.c and essence6.c. The main difference
> is that essence6 is sensitive to line breaks. It's been several weeks
> since I looked at this stuff, I sure hope I'm sending you the correct
> files! I did most of my testing with essence6. I have 9 different
> functions I played with, so there's some room for confusion.
>
> You'll need to get the gnu arbitrary percision math lib, libgmp from
> your favorite gnu server.
>
> I'd be happy to put one or both of these sums into a form that can be
> included into the dcc project if there's interest. At the moment, they
> read stdin and print out a relatively long number on stdout. Try adding
> or deleting a line from the imput file and running it again. The output
> should be the same or very similar down to nearly the final digits.
> Obviously these fuzzy sums work best with large files and few mods. The
> good news is that you can always chop the number at a certain number of
> digits to make them fuzzier. You'll have to do that anyway since dns
> limits the overall length of a name which can be searched for. Also
> need to compact the number to use all available bits. I'm perfectly
> willing to do this, I just haven't yet since I was first experimenting
> with fuzzy functions first.
>
> Let me know what you think.
I'll look 'em over.
> -Mike
>
> p.s. also, please don't repost this with my email address, use
> http://www.grant.org/~mg-dcc instead, cheers.
Right. I understand.