Yes, that is why they were rejected.  They should start with zeros and if they don't have enough zeros they are rejected as something your client should never have submitted.  Now, this shouldn't happen.  For this to happen there needs to be some misunderstanding between the client and the server about what data you should be hashing.  I'm not sure what caused that.

Yeah, your proposal is a clean way of doing it.  Mine was just a hack to work within the existing protocol.

That makes sense to me.  I know the ship has long since sailed, but I kinda with these weren't json arrays but instead where key value pairs so that it would be a little more self describing and easier to add various extensions.  Maybe any extensions should be done as pairs rather than tacking values on to keep things clear?

So, I've just made SockThing use the same extranonce1 for all connections.  I am also adding an extra random number to the coinbase and generate a new coinbase for each job.

Now reconnects should work fine, except clients have no way of knowing that they will.

I'm thinking maybe adding a field to mining.subscribe that indicates this feature exists.

However, I don't imagine other pool operators are inclined to build a new coinbase for each user so I don't imagine my solution will be for everyone.

Rejects over all are very low:
+-------------+----------+
| reason      | count(*) |
+-------------+----------+
| NULL        |     3756 |
| H-not-zero  |       20 |
| quite stale |       16 |
+-------------+----------+


The last two here are from me testing the difficulty target check.  It looks like the rest of these are failing the check, as they should.  Those hashes certainly aren't right.  That means there is some mismatch of data between the client and server.

Is it possible that your client is reconnecting and trying to submit shares for jobs it learned about on a previous connection?  I don't have that working quite yet.

+------------------------------------+------------+------------+------------+---------------------+------------------------------------------------------------------+
| username                           | our_result | reason     | difficulty | time                | hash                                                             |
+------------------------------------+------------+------------+------------+---------------------+------------------------------------------------------------------+
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 07:58:52 | 743523edf408749ae13d9ee873c092f3b8edf728eee6d184bc9ce0952f6e500b |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:37:44 | 757a6e1c4113c1f7501c36bd31ce7e4e4ce4d7eb234ecab9e1602bc85dce73de |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:37:41 | 4890db9d3eef1fd2d3ebbe842460d066958021fc8c6390db149621bf5814e969 |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:37:37 | 76dc456b0b1636ccdfa32f8baf1ad96474c283ad5a991ff40a45a6ea7947b5ac |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:37:35 | 6faecf3b91bde22c912c1406918beb3838f1ae148ebbacfc83a5e1e7347295fd |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:37:22 | 9384c003d606b77f46ad474d87d74e693e709d6a9d8e1c1e23d08d906ee11cbd |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:37:14 | c5def3f1a5b47dfc9d9219f5192a4c62a5d6c7fe7502d5eca1c3cf721ce2925d |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:37:12 | c47bec16ae761e326d5f57ecce3d720bea977ec26ea9d21664280ffb314c9bb2 |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:37:12 | c50de5de26c86d2227bf1d15df1a4b8f096eb0b1bd7fb8b78ac4a2739f49dd82 |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:37:01 | 0027b7553f825aa14bb188f5a034d172eab22ce7802adc63f96efc103d9a182f |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:36:52 | af0ca2e7217a48cae672e05318dbae53b1ad1cdc7be87c57826215269bef2943 |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:36:45 | 231392a762f6c4b1551e39e6b5b64407b4d956024418df2db403fe65289502c3 |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:36:39 | 3170f121c3591406476deacc8eb8c5f0c9f4a227e72ed7f1509e632c5ec68de8 |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 05:36:39 | a9692eb13e11e2a2b9bf6efb4b9cc1a6fe4d8d03ac847e61e91759065020e678 |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 02:24:45 | 79627670370a6c3670fabed26eb5537e7634544a1e397f4febd32f2d33b97d50 |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 02:21:42 | 80aedaae2514553fcdee97296def3670d2ef66d44d24ffe3b5c7da729fb61709 |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 01:24:01 | 8e8dc6c6354d3def9061cefdebdbb70622528969e57eb1dda03b86273ba7efa2 |
| 139Ez4YdvCU1dgbXVqAcsB62zzTdRMA43g | N          | H-not-zero |         64 | 2013-02-14 01:15:08 | 727be5db1bb10faa88f7950c837a62d7e811967d3c58ad92e91e609c012eff27 |
| 1FdKhHttWiwq6Wip1BBJG1f6ffD8pWYFiE | N          | H-not-zero |          4 | 2013-02-13 20:11:48 | 0000000027cb6379960f461326a676f1493d074803b0172ae3ca7e05b6d63410 |
| 1FdKhHttWiwq6Wip1BBJG1f6ffD8pWYFiE | N          | H-not-zero |          4 | 2013-02-13 20:11:38 | 000000001e9b9a1470e322aaac34201678b1f42244e29d39b794b8b6241cd964 |
+------------------------------------+------------+------------+------------+---------------------+------------------------------------------------------------------+

I've put the code on github:

https://github.com/fireduck64/SockThing

I've put my code on github in case anyone is interested:
https://github.com/fireduck64/SockThing

It seems to work but is certainly not done.  I need to add metrics, probably some job management and clean up the multi-connection same user story a bit.

I don't know what the issue is, but I can't get 2.9.5 to work either.  Later versions seem to be fine.

I have made a java stratum pool server implementation.  It seems to be working pretty well.  Once I pull my passwords and settings out of the source into a config file I'll put it up on github.

Questions:

1) Should a work unit (from a mining.notify message) be usable by a client if the TCP connection breaks and they reconnect?  It makes sense to me that it should, but I am having trouble because the mining.subscribe seems to occur before the mining.authorize so I need to issue (or reissue) the same extranonce1 before I know which user it is.

2) I should be able to support a bunch of authenticated workers on a single TCP connection, right?  This means they are all going to work with the same mining.notify messages and with the same extranonce1 values?

If I do this, it makes me dedup code a little more complex.  If I don't do it, it probably disables some interesting proxy use cases.

I could also see an issue with select-able difficulty pools (like mine) where a user creates a stratum connection with a bunch of workers at different difficulty levels and then when they find a share they submit it as whichever worker has the highest difficulty that it matches, thus greatly increasing their pay at my expense.

Yeah, I was restarting it to try to improve something and it didn't work so well.  It is back up now.

I am impatient, I've put in the redirect header and we shall see what happens.

If anyone wants to try it out, it is operational:

stratum+tcp://stratum.hhtt.1209k.com:3333/

I'd like to run it for a few days with some test users and then I'll put that header in the pushpool.

Ok, mined some blocks on testnet3.  http://blockexplorer.com/testnet/address/mjo9y54EJthcxpZc4pLGHWvf4n8E3vaTrN

Since I am making this terrible coinbase transaction myself, I'm thinking of doing something fun like half of the transaction fees to miner who found the block right from the coinbase.  Since I am building the transaction per job that isn't really much trouble. 

Probably.  I'm just glad I have the miner side working, that way I can validate shares submitted by cgminer and assuming cgminer is correct I'll know if the rest of my nonsense is correct or not.  So I should have a solid idea before I even try making some blocks on testnet.

With some help with Mr. Hearn over on the bitcoinj mailing list I think I have it figured out.  I also have a grip on the merkle branches sent over the wire with the stratum jobs.  I hope to have the entire thing working on testnet tomorrow or the day after.  I expect tons of silly endian problems.

After that, I'll do some cleanup and make it configurable and put it on github.

Should be fixed now.  Will be reflected on the web page in a little bit.

Thanks for pointing this out.  I think I know what happened.  I was updating a bitcoind on my backend.  It switched to level db and was downloading the blockchain.  That is all good, except it is used to determine the difficulty for the purpose of share payments.  It probably ran a number of times while the blockchain was being downloaded and got some incorrect difficulties in the process.

I've stopped payments for a little while I cleanup the DB.  Some people will probably have been paid too much and go negative for a while (maybe forever if they stop mining).  But I won't ask people to be responsible for my mistake.

Sweet.  I am working on my own stratum mining pool implementation in java.  I have the stratum side working, which includes getting work to miners and processing the shares.  I am working on the bitcoin side of actually making the blocks.

If anyone has an pointers on making a two part coinbase transaction in Java (in which I can cram my 8 bytes of extra-nonce), I'd love to hear about it.

I had my fiance read about stratum to me in the car on the way home yesterday.  She wasn't very pleased about that.  Anyways, it sounds like the time rolling wasn't as flexible as I thought it was.

I'm looking into spinning up a stratum node now.
 

I expect they will work fine just cranking the difficulty.  From what I've read, the Avalon units are based on a recent cgminer and BFL units (at least the Singles and hot peppers) will be USB so the user can use existing mining software (like cgminer).  So I don't expect any problems.

If there are problems, I'll be happy to work with miners to solve them.  I don't support Stratum yet but if that ends up being needed or a good idea I can see doing that work.

I should really check my changes to pushpool into a branch and sync it up to the head sometime soon.