Come-from-Beyond
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
January 03, 2014, 10:27:02 AM |
|
Adding 1-4 bytes won't hurt much.
I disagree. We r moving to 30.000 tps goal, every ounce of performance is important.
|
|
|
|
Come-from-Beyond
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
January 03, 2014, 10:28:23 AM |
|
So, you seem to assume that the cause of the problem can not be in the core?
Right. Code that sends money is very simple.
|
|
|
|
mcjavar
|
|
January 03, 2014, 10:29:25 AM |
|
So, you seem to assume that the cause of the problem can not be in the core?
Right. Code that sends money is very simple. Although there will be a new client soon (hopefully) such a bug shouldn´t be left in the current client. Can´t you try to debug it CfB?
|
|
|
|
NxtChg
|
|
January 03, 2014, 10:31:32 AM |
|
Adding 1-4 bytes won't hurt much.
I disagree. We r moving to 30.000 tps goal, every ounce of performance is important. Oh, please! 8 bits out of 256-bit address? 3% increase. There's probably much more bloat than this already in the system. It can even be 4 bits. Anything. A single party bit is better than nothing. The NXT system will not be treated seriously if it can send money to an arbitrary number without any checksum at all and this money will be locked forever! You simply won't havee a chance to reach 30.000 tps with this design
|
|
|
|
Come-from-Beyond
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
January 03, 2014, 10:31:49 AM |
|
Although there will be a new client soon (hopefully) such a bug shouldn´t be left in the current client. Can´t you try to debug it CfB?
I would if I knew how to reproduce this situation.
|
|
|
|
landomata
Legendary
Offline
Activity: 2184
Merit: 1000
|
|
January 03, 2014, 10:37:56 AM |
|
sorry guys but I have to go ...will be back soon....Nxters are making history today
|
|
|
|
ImmortAlex
|
|
January 03, 2014, 10:39:37 AM |
|
Just a friendly reminder... There will be a Bitcoin Conference in February held in Berlin. http://www.mediabistro.com/insidebitcoins/I am not from Berlin, but would love to represent Nxt on this venue. I decided to take 3 days off and fly over to Germany and use the opportunity to spread the word, let others know about Nxt and it´s features. I am sure that we all will benefit from having someone around at a venue like this. I am asking the community to support these efforts by donating to help fund the journey and the purchase of marketing materials like T-Shirts & pendrives with Nxt logo/slogan on them, printing some cool Nxt bills, and so on. I will document all the expenses and will make the spending of the fund transparent. I will also broadcast from the venue and post some pictures I received donations from 3 persons so far... I am asking community members, escpeially with big stakes to support this and help with funds AND ideas how to make Germany aware of Nxt. Even a couple of hundred Nxts will help. 11433600460445633305Thank you! I strongly consider that support your presentation is much better then spend coins to all this "please-give-me" 's I see in this thread. And I wonder why your request has almost no reaction among Nxt holders...
|
|
|
|
NxtChg
|
|
January 03, 2014, 10:42:44 AM |
|
Right. Code that sends money is very simple.
So it's 100% prone to memory corruption? Come on...
|
|
|
|
notsoshifty
|
|
January 03, 2014, 10:42:59 AM |
|
There must be CRC added into addresses ASAP!
I agree. I thought about this recently and not found any reasons why we can't add the "check digit" to verify the account (as example GS-1, IBAN, etc). Use Alias System. With this talk about sending to alias, do you have any thoughts on my question about what/when does the alias lookup? As a side question: one important feature for nxt is the ability to send funds to a named alias rather than a number. I can see this being implemented in two ways:
1. Purely client side. The client looks up the alias to get the account ID, and then creates a transaction with that account (not alias) as the recipient 2. Within network/API. The client creates a transaction with the alias as the recipient, and the network (i.e. forging node, and other nodes through validation) translates the alias into an account id at time of forging
The second seems to me to be cleaner and preferred. Example: a transaction might take hours to get processed, and what's important is the account that the alias corresponds to at the time it gets processed, not when it is initially shoved into the network.
However, this would require API changes. And so any URI conventions that are adopted by browser extensions/addons would need to be adopted into the API.
Another point in favour of #2: I would rather see in the block chain a transaction that says I sent to 'bob', not to '123456789'.
|
|
|
|
mcjavar
|
|
January 03, 2014, 10:44:41 AM |
|
Just a friendly reminder... There will be a Bitcoin Conference in February held in Berlin. http://www.mediabistro.com/insidebitcoins/I am not from Berlin, but would love to represent Nxt on this venue. I decided to take 3 days off and fly over to Germany and use the opportunity to spread the word, let others know about Nxt and it´s features. I am sure that we all will benefit from having someone around at a venue like this. I am asking the community to support these efforts by donating to help fund the journey and the purchase of marketing materials like T-Shirts & pendrives with Nxt logo/slogan on them, printing some cool Nxt bills, and so on. I will document all the expenses and will make the spending of the fund transparent. I will also broadcast from the venue and post some pictures I received donations from 3 persons so far... I am asking community members, escpeially with big stakes to support this and help with funds AND ideas how to make Germany aware of Nxt. Even a couple of hundred Nxts will help. 11433600460445633305Thank you! I strongly consider that support your presentation is much better then spend coins to all this "please-give-me" 's I see in this thread. And I wonder why your request has almost no reaction among Nxt holders... You are the only one who is replying when I write about the conference. I really don´t know why people are so ignorant regarding this. Anyway, I won´t give up
|
|
|
|
intel
Member
Offline
Activity: 98
Merit: 10
|
|
January 03, 2014, 10:45:07 AM |
|
Although there will be a new client soon (hopefully) such a bug shouldn´t be left in the current client. Can´t you try to debug it CfB?
I would if I knew how to reproduce this situation. I bet i know the real reason for this. I had this issue also (transaction was sent to another account). So, after i clicked "send", the GUI waited too long. Finally it said "sent". Then i checked blocks tab and found wrong recent blocks number, unconfirmed transactions and other signs of freezed java process. I instantly logged into the server and found several exceptions (not related to timeout) in the log. Few minutes later daemon died by not enough memory exception. Instead of that useless CRC checks, it's better first to implement an additional check that the process itself is heathly and can be used to transfer money. Right, just few integrity checks before NXT transfer execution. It's Java. It's cross-platform. It's terrible and not really your friend. Right. Code that sends money is very simple.
So it's 100% prone to memory corruption? Come on... Yes.
|
|
|
|
wesleyh
|
|
January 03, 2014, 10:46:32 AM |
|
@CFB - I'm using the api to send money (via my nxtra faucet) - I get back a transaction ID. However several times this transaction doesn't actually happen. How can I know the transaction failed or do I need another approach?
Just rebroadcast the transaction if it's not confirmed (broadcastTransaction API call). In most cases transactions r ignored coz of incorrect time. Adjust computer clock, only +15 sec in the future is allowed. Actually the time is correct, the problem corrected after I restarted the java client. Any ideas why that is? It was on the correct block before.
|
|
|
|
Come-from-Beyond
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
January 03, 2014, 10:48:09 AM |
|
Right. Code that sends money is very simple.
So it's 100% prone to memory corruption? Come on... What do u mean "memory corruption"?
|
|
|
|
exokk
|
|
January 03, 2014, 10:48:34 AM |
|
I cannot understand why someone would use java for such a critical system. Stick with C/C++ (well, c# would be ok, too)
|
|
|
|
Come-from-Beyond
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
January 03, 2014, 10:49:10 AM |
|
There must be CRC added into addresses ASAP!
I agree. I thought about this recently and not found any reasons why we can't add the "check digit" to verify the account (as example GS-1, IBAN, etc). Use Alias System. With this talk about sending to alias, do you have any thoughts on my question about what/when does the alias lookup? As a side question: one important feature for nxt is the ability to send funds to a named alias rather than a number. I can see this being implemented in two ways:
1. Purely client side. The client looks up the alias to get the account ID, and then creates a transaction with that account (not alias) as the recipient 2. Within network/API. The client creates a transaction with the alias as the recipient, and the network (i.e. forging node, and other nodes through validation) translates the alias into an account id at time of forging
The second seems to me to be cleaner and preferred. Example: a transaction might take hours to get processed, and what's important is the account that the alias corresponds to at the time it gets processed, not when it is initially shoved into the network.
However, this would require API changes. And so any URI conventions that are adopted by browser extensions/addons would need to be adopted into the API.
Another point in favour of #2: I would rather see in the block chain a transaction that says I sent to 'bob', not to '123456789'. "Send to alias" payment should solve this, right?
|
|
|
|
NxtChg
|
|
January 03, 2014, 10:49:58 AM |
|
Instead of that useless CRC checks, it's better first to implement an additional check that the process itself is heathly and can be used to transfer money.
No! The CRC check is precisely what would protect from an unhealthy node, because other nodes would simply reject the transaction! It's not feasible to self-verify that running code is correct.
|
|
|
|
Come-from-Beyond
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
January 03, 2014, 10:54:15 AM |
|
I bet i know the real reason for this. I had this issue also (transaction was sent to another account).
So, after i clicked "send", the GUI waited too long. Finally it said "sent". Then i checked blocks tab and found wrong recent blocks number, unconfirmed transactions and other signs of freezed java process. I instantly logged into the server and found several exceptions (not related to timeout) in the log. Few minutes later daemon died by not enough memory exception.
Instead of that useless CRC checks, it's better first to implement an additional check that the process itself is heathly and can be used to transfer money. Right, just few integrity checks before NXT transfer execution. It's Java. It's cross-platform. It's terrible and not really your friend.
GUI waits until all peers receive the transaction, the time heavily depends on pushThreshold value. Before sending a transaction NRS checks that its signature matches signed data, if recipient value was changed after signing the transaction wouldn't be confirmed.
|
|
|
|
Damelon
Legendary
Offline
Activity: 1092
Merit: 1010
|
|
January 03, 2014, 10:54:20 AM |
|
Just a friendly reminder... There will be a Bitcoin Conference in February held in Berlin. http://www.mediabistro.com/insidebitcoins/I am not from Berlin, but would love to represent Nxt on this venue. I decided to take 3 days off and fly over to Germany and use the opportunity to spread the word, let others know about Nxt and it´s features. I am sure that we all will benefit from having someone around at a venue like this. I am asking the community to support these efforts by donating to help fund the journey and the purchase of marketing materials like T-Shirts & pendrives with Nxt logo/slogan on them, printing some cool Nxt bills, and so on. I will document all the expenses and will make the spending of the fund transparent. I will also broadcast from the venue and post some pictures I received donations from 3 persons so far... I am asking community members, escpeially with big stakes to support this and help with funds AND ideas how to make Germany aware of Nxt. Even a couple of hundred Nxts will help. 11433600460445633305Thank you! I strongly consider that support your presentation is much better then spend coins to all this "please-give-me" 's I see in this thread. And I wonder why your request has almost no reaction among Nxt holders... You are the only one who is replying when I write about the conference. I really don´t know why people are so ignorant regarding this. Anyway, I won´t give up I just heard I am going to represent my company on the Ambiente fair in the Frankfurter Messe in February. My main work will of course be for my own company, but in the five days I am there I will meet loads of business people (wholesalers and big retailers) from all over the world. I am certainly willing, if it doesn't conflict with my business, to introduce these people to Nxt. I would need some nice marketing tools though, to at least get them hooked/interested. We're in te beginning of development, so it's natural that we can't give them totally finished tools to show, but anything helps at this moment to at least BE there. Can we figure something out to support these things, so people who are actually promoting Nxt in the real world have the materials, a good website (marketing website, not a technical one, so a presentation site), and factsheets and goodies to hand out? Maybe we can pool resources as marketeers.
|
|
|
|
Come-from-Beyond
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
January 03, 2014, 10:55:07 AM |
|
Actually the time is correct, the problem corrected after I restarted the java client. Any ideas why that is? It was on the correct block before.
What deadline do u use?
|
|
|
|
NxtChg
|
|
January 03, 2014, 10:56:33 AM |
|
What do u mean "memory corruption"?
Well, I mean precisely that. Corrupted memory. I don't know what more to say... Surely, as a programmer, you now what I mean. You cannot guarantee that your code is 100% prone to arbitrary changes of any of its memory bytes.
|
|
|
|
|