5-man TBs, limited bandwidth, and Hyatt's FTP site

[ryan_hirst]

Thought this was worth a new topic:

It seems to me the "EGTB Online" site should first direct 5-man TB seekers to Hyatt's FTP site rather than to "practice using emule and download them by these links".

Hyatt's site has all 3-, 4-, and 5-man TBs at spectacularly fast download rates. I just logged in as a test and got over 200KB/s.

5-man TBs are available in a number of locations. Shouldn't those be recommended FIRST? Why not offload much-needed bandwidth whenever possible, freeing it up for the 6-man transactions, files which can be found literally nowhere else?


That way:
1. The people who need the 5-man sets will get them MUCH faster.
2. Files for which this community is the only source will receive the maximum possible number of transactions between users (without actually refusing to host 3-5 man TBs, which would be silly).

Thoughts?

[Alkelele]

There are both pro and cons. I understand the speed issue and freeing up bandwith.

On the other hand, I personally decided just to go along and start from scratch with Emule, even though I probably have some 3-4-5 tb's on my Fritz dvd. Thing is, this was just a comfortable choice, not needing to bother about compatility issues like the one you mentioned in another posting, or any other such manual thing.

I think that it would be a good selling point for this project if it was simply the place to go to get ANY tb's, 6 piece or not. Then more people might get sucked in and start sharing 6 piece tb's eventually.

Just a point to consider.

[ryan_hirst]

Seeing that most people at ChessBase don't extensively use 6-man sets, I was worried that most people routed here for the 5-man set would lose interest and disappear, placing an unusually high load on those files from all parties... without the long-term return.
The more I think about it, though, I think you're right. We need people getting sucked in. The eMule help page lists 100 or more sources as well represented. Hehehehehe. Here it's 3-10+. Also, even a high-dropout-rate influx is still an influx. Anyone who stays, helps. It still still seem worth mentioning on the main page that 5-man sets can be downloaded way faster on the Crafty site. Maybe it could say:
1. Go download eMule and get started. Pick something useful to download like KRP KRP. Ok, now while you're waiting,
2. Go to the Crafty ftp site (ftp://ftp.cis.uab.edu/pub/hyatt/TB/3-4-5/) if you still need any 3-5 man sets.

[guyhaw]

I decided to make sure I had all 3-to-5-man EGTs ...

... so that I could create a Data Assurance Certificate for them with manifestly-correct MD5sum and eDonkey-'MD4' checksums followed by an 'integrity ok' message from datacomp.

I used KK's home page to request whole tranches of files, started eMule and then manually downloaded from Hyatt those that had not started eMuling.

So I got my eMule training, learned quite a bit, and fast bandwidth at the same time.

g

[Kirill Kryukov]

Hi all,

I was thinking to link to existing sites with 5-men download, but then I though eMule availability is good enough. My reasons are mostly same with Alkelele's:

1. Everything is available at the same site, you don't have to hunt for those 5-men files here and there. It's good image too to have everything at one place.
2. It sucks people in.
3. 5-men bases are tiny compared to 6-men. There are 6-men sets larger than the whole 3-4-5-men collection. And it's really better to learn eMule with something little.
4. All 5-men bases are reliably shared on ed2k network by now. You will get them reasonably fast anyway.
5. eMule checks the file integrity with MD4-like digest, FTP does not do that so the chance to get corrupted file by FTP is larger.
6. FTP is up today and down the other day. p2p is forever.

[Arkansaw]

I just want to concur that Hyatt's ftp is indeed blazing fast...it only took an afternoon to leech 4+ gig of data (I already generated the rest myself) at roughly 200kb/s.

Provided that one can achieve a similiar connection speed and such behaviour won't result in too much excessive load on the ftp, leeching the 5-man bases from there is the best solution I believe

[Kirill Kryukov]

OK, I think it does not hurt to mention it, so I added link to Hyatt's FTP to the project page. I also added a couple of links to eMule help.. Please let me know if we are still missing something there. Thanks!

[Ray]

OK, I think it does not hurt to mention it, so I added link to Hyatt's FTP to the project page. I also added a couple of links to eMule help.. Please let me know if we are still missing something there. Thanks!

Where is the link to your page Kirill ? I don't see it in the welcome post

[Kirill Kryukov]

Hi Ray, the page is here (also listed in this thread). I'll add it to my sig.

Edit: Added. Now I'll spam it with all my posts.

[gambit3]

just a note. emule will not deliver all 299 3-5 man tables. i have the complete set, and emule keeps not sharing 5 of them (kbk black, knk black, krkr black kqkq black and krkp black) because it considers them duplicates. in sense of content, emule is right - byte for byte, these files are identical to their white counterparts, however they (particularly krkp) should not be overlooked in a download plan. hence there must be a link to hyatt for AT LEAST 3 and 4 man bases.

so far, i haven't seen this same problem for any 5 or 6 man bases...

[Kirill Kryukov]

I thought the "nbb" part of symmetric sets is not necessary at all? All engines that I tried are happy without it, they just use "nbw" part for both white and black. Do you know any engine that particularly requires having both files for symmetrical sets, even though they are identical?

[gambit3]

I thought the "nbb" part of symmetric sets is not necessary at all? All engines that I tried are happy without it, they just use "nbw" part for both white and black. Do you know any engine that particularly requires having both files for symmetrical sets, even though they are identical?

this is, in fact, true for ALL, not just symmetric, sets. however, for the purposes of generation, both are required. as mentioned in other places, i use a only a black (nbb) set for tournament, etc., play, allowing me to use 3v3p within the confines of a single 200G drive. i keep the rest around (what i have of them) for generation purposes, and because i am so familiar with the pros and cons, i generated the bishop and/or knight 6 manners before coming to emule for the rest. thus i am certain of the integrity of my 5 man and bishop and/or knight (so, not pawn, queen or rook) 6 manners. for testing purposes, i have run tbgen -q over them once (don't need any checksum for this, so i can just go on the hash that people tell me is correct for emule and not download the ones with other hashes, and this way is very quick - 57 disc accesses maximum i have seen) for a small degree of security.

i can vouch for the hyatt ftp being >1.2M/s too. gateway to gateway transfers are to drool over. it took me just under 2 hours for complete 3-5 from hyatt site, so leave emule for the 6s...

how many topics can i put in one reply? :>

ok, a further discussion of what really happens with engines with partial tb sets is here. you'll have to scroll down (about half way) to my entry, but it is a long entry, so i'll point to rather than quote it. space and all that...

good luck and happy hunting.

[Wilfried Eberl]

I thought the "nbb" part of symmetric sets is not necessary at all? All engines that I tried are happy without it, they just use "nbw" part for both white and black. Do you know any engine that particularly requires having both files for symmetrical sets, even though they are identical?

Even if you generate the EGTB yourselves, the "nbb" part of symmetric sets will not be produced. The generator just ignores them. This simply means: The mentioned "nbb" files are not neccessary!

I did generate all the 5-men an few years ago and therefore i experienced the generator's behaviour mentioned above and none of my engines do miss these files.

Kind regards,
Wilfried

[gambit3]

interesting, since i have 299, and that includes all, and that was also from generation. with tbgen, the original recompiled to allow pawned 5 man. and the white and black are produced together, not separately. at the same time even. and only if both subsets are present. try again, this time recheck. try removing just one of these files and try to regenerate kppkp. see what happens. i can assure you, it will not be an immediate return to command prompt, no matter how fast it seems. in fact, it's a pity you have to compress the files since .c is not supported at this forum, or i would add the source that produced it, but that same source is available other places anyway. please make sure you have counted right before giving an answer.

[phareast]

interesting, since i have 299, and that includes all, and that was also from generation. with tbgen, the original recompiled to allow pawned 5 man. and the white and black are produced together, not separately. at the same time even. and only if both subsets are present. try again, this time recheck. try removing just one of these files and try to regenerate kppkp. see what happens. i can assure you, it will not be an immediate return to command prompt, no matter how fast it seems. in fact, it's a pity you have to compress the files since .c is not supported at this forum, or i would add the source that produced it, but that same source is available other places anyway. please make sure you have counted right before giving an answer.

You quote 299 as the number of files for 3-5 man, but Hyatt has only 290 files of 3-5 man tables.

-p

[gambit3]

please recheck. i didn't say 299 for nothing, and i'm still busy reading through a lot of stuff at the reading math department... i urge a correct count before proceeding. generator gave 299, and it is the same one nalimov wrote.