I’ve no sense of it but wonder that passing around vault loads of data to new vaults in the first few days, is a lot of traffic and once those settle it’s going to be quicker. Also, the browser I wonder still caches, so if you are returning to the same page, you might not be downloading anything new. A quick look sees vault suggesting size of only 150 vaults, so not much difference there.
Today safe://constellations/ takes about 5s to load, as opposed to about 10s yesterday. I’ve done a hard refresh too. Also the network (my section of it anyway) seems a lot more stable, with none of the rapid churn events going on. The only negative thing I can spot is that the flying squirrels safe://wingsuit.draw stop flying at 24 seconds - maybe the data-loss problem?
328 get request for every chunk stored. Somebody is uploading really populair data or intentionally slowing down the network. 38MB stored log file size 58MB
328 get request for every chunk stored.
Somebody is uploading really populair data or intentionally slowing down the network.
38MB stored log file size 58MB
Was there some suggestion that popular files become quicker as the network organizes itself?.. I can’t think how that is done but perhaps constellations is among the more popular files atm.
Ha - I should have kept quiet. Logged out and can’t get back in again now!
Edit. And now my vault has crashed.
Will we see another test before Alpha 2?
Hi, great update!!! I’m sorry to ask again, any plan to enable the Nat traversal?
Not problems for me this time. I can log in in my account. However, look like somebody try to bother the network downloading a lot of data. In the last hours the GETs are increasing a lot.
I have the same issue
Test 15 was very stable but now people say test 16 isn’t? What’s going on. Did the code get worse somehow recently? Need alpha 2!
We need Alpha 2 to be stable it goes out to the masses. I’d rather have several more tests than screw that up.
I guess they know what changed between 15 and 16 so easy fix… fingers crossed.
During my first vault connection there was a lot of errors. The log file grew rapidly. Had to terminate it. Not sure if it was something on my end causing it. The following connections seem to work ok.
Seems like there has been further data loss - SEEMS like - Maybe I am wrong…
Can anyone access safe://rust,scotcoin?
I can’t access that site after I had to make a new account.
also I got a time out trying to upload a small site with my new account.
With Test 15 someone overwhelmed the network by filling it up. Now you say someone downloads aggressively.
It has to be the same person right?
Caching is only implemented for Immutable Data, but safe://constellations is entirely in Structured Data (with files under 3KB and directories under 100KB), and any mutable data cannot be cached because it is not possible to mutate instantaneously every cached copies in the network.
This is true for SD but also for AD and MD.
If it is, then shouldn’t we know or have some idea of who that is, as invitation only?
You don’t an account to access the network, so no.
So what would the fix for this be?
I’m thinking Safecoin won’t fix this one as free gets and requiring an account isn’t a long term fix?
There are several things we have looked into, but basically we are being forced to implement things we did not wish to until after Alpha 2. It’s a bit of a pain but also a good thing in a way. As we tell people what limitations are and help folk go through the code then it seems the bad side of that is that before the system is ready for such attacks we have at least one person trying to cause a nuisance. It’s not clever but probably is not too bad, just we need to actually treat early tests with a bit more care in terms of not allowing such openness.
We know the next steps and vaults will require some additional parts now and we cannot just release the next part for everyone to test, as quickly. Instead we must now keep the releases back a bit and make sure we add in the components to prevent such attacks, like start a bunch of vaults, switch off / on, disconnect partially off and on etc. It’s not clever and easy to do, we let folk know it’s easy, but players will play and all that jazz. So we will have a wee bit of work to do.
Clients are partially protected with tokens, but what a shame really, it’s life though and it does not worry us at all, just a wee reminder that not everyone is trying to help, some (very very few) will just do what they can to spoil a party, however they wont spoil this party
In terms of what is really required to sort this completely, then it’s node age, secure name, full sig checks, safecoin, data chains etc. So the complete answer is when everything is in place this wont happen, but we really must work more now to put in place temp parts of code to allow us to iterate in the open while we do implement all those parts. We can though and we will, we will just make all the good folks happy and give even fewer, not so supportive, a game to play.
[Edit} I should add, we will need to be careful with community networks now as well, if these are open then they are an easy target to spoil.
MaidSafe Dev Update - April 13, 2017