That´s why I asked to consider that we are miles away from a fully functional version. If you want something you need to go with what you have. (Sorry, if I sound too pragmatic, it´s just the way I am)
1 Like
I think he would’ve been highly satisfied with Test 2, though it didn’t last very long. Would’ve been a good way to learn about the community test, too. Can’t blame for not wanting to try the community test 3 since he might end up with file drops again… call it being unlucky. (Personally I’ve been holding off on the community one until later tests. Also I was very patient with test3, highly enjoyed the pay-off of being able to eventually/finally upload files again after expected hiccups. Different strokes.)
1 Like
I don’t mind it failing in fits and jolts but if it’s going to do that I’d like to be able to log in again after it fails. I don’t want to have to keep creating new accounts. Can’t test much without a stable account.
1 Like
Then you’ll have to wait for a later version.
On the test nets be prepared to cycle through several accounts if you participate for more than one day.
However, in my experience, if you access your account and its data every hour or so then it will stay alive.
1 Like
4 Likes
In the context of @Blindsite2k’s complaints about account creation: I had the impression sometimes that the community1 network somehow still remembers “part” of my account after some ours. I could not log in to my account, but I could not create an account with exactly the same credentials either. A small variation in the credentials immediately allowed me to create a new account.
2 Likes
Yes, this was expected. The test was not focused on keeping all chunks stored, it was focused on keeping the routing tables alive. an account consists of chunks as well. David said next test will have both optimized routing and reliable storage of chunks.
5 Likes
Not remembering your accounts is one aspect of failing.
And many changes in the kademlia_routing_table too:
Very interesting to make the group size configurable. I imagine that will significantly reduce churn.
2 Likes
When they started the most recent testnet (3) the devs specifically said that it was to test routing, churn, etc and that data, uploading and account creation would be spotty at best. This was intentional and I fully expect to see the next version purring like a kitten… 
4 Likes
That would be likely to break compatibility of different versions of the vault, that won’t be able to coexist on the same network. With many other kinds of changes, such as log verbosity or aggressiveness of crust, or efficiency of data storage, they probably can (mostly) coexist.
The devs progress a lot in past few months. The last three test was impressive feat.
I cannot wait for tomorrows update.
8 Likes
Yes, test3 gave us access to a deeper bug in holding the kademlia invariant that we have been chasing. 1 & 2 were to test mainly network handling and we pushed then to move blocking IO with threads to a non blocking OS based model (mio). Work there moving very fast.
The early client tests against our static network (no churn) were to prove account creation web building etc… So we know that all worked fine. The latest tests were specifically to kill parts of the network. The last test for instance was only 100 nodes to kill it faster and allow us to look deeper into an issue we wanted to confirm. It worked, we “think” we have it now, but still testing to make sure.
We may do another test quickly, but evaluating whether to hold a few days to also include the newer async networking in crust. So we are looking now at what the next test rollout will be and if not tomorrow then when, but the guys in crust are really under a load of pressure to get there soon. We will see though.
We all want to be released back into vaults and client side to get the API’s there moving along much quicker, but it’s all a matter of small team, big job. We will allocate and move resources a lot in the next short period though 
31 Likes
Amen. From the outside looking in it appeared there was a significant effort on visualizing logs (as opposed to just some simple rotated file for now). I personally have halted all development somewhat knowing that the launcher API may have upcoming backwards-incompatible alterations (or at least I hope so around how the encryption is handled today). I am also looking forward to discussion resuming on Discuss - 0028-launcher-low-level-api · Issue #115 · maidsafe/rfcs · GitHub specifically about how data map serialization will work.
2 Likes
Yes, I agree. My observation in particular was that while the account itself gets deleted/lost, some kind of hash or similar, derived from the credentials, seems to survive, with the result that I can not create a new account with the same credentials (my pure speculation/empirical observation). I discovered that @bluebird posted a similar observation ~ 24 h ago in the community1 thread.
I have created the exact same account on every test net, with same PIN, Keyword, and password. No problems. On Test Net 3, I created that same account three or four times, as the network lost my info (which was the expected behavior of Test Net 3) and I had to recreate it.
So for me, there has definitely not been any data residue between test nets.
I have found that if you wait long enough then the residue disappears as well and a new account can be created with the same credentials. But that residue, whatever it is, seems to be the last type of data to disappear.
4 Likes
I see, so maybe I was either temporarily not able to log into my old account, or to create a new one (which worked seconds later with different credentials)
Well, it’s risky to generalize from one or two instances. On my “to do list” is to develop some metrics for these kinds of failures, so we can make clear statements about it.
1 Like
Yes @Blindsite2k I’m right there with you, totally support you on these views. We’ve been here forever and it will be so powerful to have a running, reliable community net with reliable storage that we can use, explore, and run test apps on. It’s been years and seeing that and having it running 24/7 would feel utterly amazing.
This could be something separate perhaps? Running independent of the very specific MaidSafe test nets that are designed to test very specific aspects of the network
This is amazing news !!
3 Likes