Looking for testers

[tombc]

i'm going to release (under gpl) beta version of a client in few days

windows alpha version: http://www.bitdollar.org/bitdollar.zip

it should be stable enought but dont put many coins on it

[tcatm]

Does it support testnet?

[xf2_org]

Will users trust a client from an unknown person, lacking source code?

[jimbo77]

What are the features?

[tombc]

it's doesn't support testnet but could easily

source code out soon (dependencies: qt, openssl), was just throwing a fast release, it can still be tested in VM

features are mainly:
-flexible wallet management
-choosing inputs and change addresses when sending money
-a small block explorer

main window view still need some improvements

i forgot to point to not try to solve block with it, it's works but it's not optimized at all
(it is implemented for an easier custom miners integration)

[tcatm]

well, if you want testers, add testnet support. That's what testnet is for.

[jimbo77]

it's doesn't support testnet but could easily

source code out soon (dependencies: qt, openssl), was just throwing a fast release, it can still be tested in VM

features are mainly:
-flexible wallet management
-choosing inputs and change addresses when sending money
-a small block explorer

main window view still need some improvements

i forgot to point to not try to solve block with it, it's works but it's not optimized at all
(it is implemented for an easier custom miners integration)

Does it support wallet encryption or just exporting/importing wallet

[tombc]

yes, there is a "set password" button in the wallet manager

wallet file format is very different from bitcoin format, the only way to move
btc between bc wallet and bd wallet is to send money

[jimbo77]

Is this https://en.bitcoin.it/wiki/QBitcoin? or totally different development?

[tombc]

no, something different, i prefered to rewrite all from scratch

[Mike Hearn]

Wow, that's a huge amount of work. Does it have many unit tests?

[tombc]

i have release source code, http://www.bitdollar.org/forum/viewtopic.php?f=3&t=2

[Mike Hearn]

Very cool indeed. Is there a todo list somewhere? I see that scripts are not handled. Are you planning a full node or a client-only node?

[tombc]

i think script support is the only thing missing to be fully compatible with bitcoin client
i'll add it (that was my last thing "todo") but that kind of scripting is quite dangerous

i've noticed two small bugs:
-when reducing program in systray, no icon appears (./res folder is missing, i'll add ressource into the program directly)
-downloading blocks at first launch can be stopped before the last block, need to disconnect/reconnect

my goal is to manage bitcoin protocol exactly like bitcoin client

[Mike Hearn]

I'd be really careful with that. At the very least try to avoid exposing mining support until you've really proven that you're following all the rules of the network (as laid out on the wiki and from the official code). Supporting scripts is not enough. Think about nLockTime, the SIGHASH flags and so on.

Really, a test suite proving you behave identically to the real client would be ideal. Satoshi was historically against re-implementations because the entire network has to be exactly synchronized and even very minor deviations can cause big problems. I don't think reimplementations can be avoided, but if you do it, be aware that you're signing up for a giant can of worms.

A blackbox test suite that can prove the correctness of both your code and Satoshis (and any future implementation) would be excellent.

[xf2_org]

Seconded what [mike] says.  For example, a client must properly handle events such a transaction being confirmed, then becomes unconfirmed due to a block chain reorg, then becomes confirmed again in a different block.

This is why I advise "normal" users to avoid all clients except mainline (bitcoin/bitcoin.git) at the present time.

No alternate client gets all these details right AFAIK, and to protect your users' bitcoins, you must.

[tombc]

rollbacks are managed

sighash is not used but it could be, i'll support it
nlocktime is ignored in bitcoin too isn't it?

i've reuploaded packages for small fixes (systray, sendmoney window size and tx are now sent correctly on the network)

[Mike Hearn]

Yes, but the code could be reactivated at some point. Your client would then not understand blocks/txns using those features and cause chain forks.

I guess my point is not "here's some missing features". It's more, how can your code build confidence that it isn't going to cause spurious (or permanent!) chain forks if it becomes popular? BitCoin itself has no test suite which is unfortunate, but its the first implementation so its behavior essentially defines the protocol. Your code has no such advantage.

[marcus_of_augustus]

How does it deal with deleted copies of unencrypted wallet left on the disk after an encryption? e.g. does it shred or otherwise wipe the unencrypted wallet.dat files?

[tombc]

wallet is always wiped and always encrypted on the disk using a default pwd "" (if no pwd set)

i'll be aware of futur bitcoin changes, it can still be used safely as simple client (well except it's a beta for the moment but it's not the same problem)
i think bd is more restrictive than bc, it should not accept a tx that is rejected by bc
so i could tests it with bitcoin client and check if it is accepted