Stable Droplet Test Network

Hi everyone, first of all thanks for the help recently with the vault/network tests. It has been really helpful. For instance knowing that NAT Traversal is as good as it currently is was a fantastic bonus and saves us a ton of work. For instance almost 60% of all dev resources went into the low level network libraries with multiple versions of connection oriented UDP and now we can move forward with TCP alone. It seems a small issue, but in terms of resources for such a small team, it is a huge deal!

However, in saying that we did find that the blocking versions of sockets in Rust were too aggressive in emulating the operating system calls. We are resolving that by again putting almost all resources into refactoring the network library to use OS native calls via a library called mio. This is great news indeed as we can then hopefully put that to bed soon and get those resources where we want them, in vaults implementing messaging/safecoin etc. For details of any of these features please see the RFC’s related to them, they should be well publicised and debated by now and how we communicate deeper features in more detail.

We broke client projects - let’s fix that now

In the pretty aggressive recent networking tests though we left app developers a little behind. There was no secure way they could use the API’s and get more involved in the RFC / API process for applications (surely one of the most important aspects of SAFE). To improve matters, there we are splitting continuing tests into Vault tests and Client tests. We will continue at least one more vault test soon, but in the meantime we will run a stable network on the Digital Ocean droplets, starting from today. This client test network will not be to test anything to do with vaults and be reserved for app developers to work against something stable.

Please download the launcher here and the demo app here

Updates for launcher include :

Updated logs to provide more useful information.
Fixes for self encryption crashing apps on network full (you probably will not see this, it is a low balance error. The client still does not report this effectively, but it will happen during the RFC process below)

Also, the Launcher is looking at some significant improvements as per this [RFC] (rfcs/0036-launcher-api-v0.5.md at master · maidsafe/rfcs · GitHub)

The other community members are more than welcome to keep participating in the vault tests, as this is a huge help (we cannot emulate humans or real world internet connectivity). At the moment we really really do not want anyone to build vaults to run against the client test network (otherwise we need to use resources to block that via changes in code).

I won’t do a dev update here, but things are moving in crust, although we have 4 of the 5 devs there on some kinds of holiday/days off next week. Although Qi is not on holiday, he is attending and presenting at a conference in Beijing (thanks @qi.ma ). There is more happening in China that we will announce very soon.

Along with all of this, we will be doing some more PR activities very soon which will enable the project to move much quicker. Again, over the next few short weeks, we will hope to have lots of information in many aspects of the project moving forward. So technical, PR, structure, resources etc. will all get a bit of a push soon.

Thanks again folks all of this testing is truly helpful and we are reacting to each and every test with a rabid ferocity to make adjustments and test again. Github issues and RFC’s are being looked at a lot, so please keep this up, no matter how small, every little thing helps to make this not only a success, but more importantly get it into people’s hands more quickly.

60 Likes

Do not underestimate the impact of China on the Safe network. If there ever was a motivated/captive audience - thems it.

Good luck.

11 Likes

Great work, I’m only with my website again.

http://polpolrene.safenet

Enjoy!

7 Likes

Seems faster… http://test.bluebird.safenet

EDIT: added http://test.bluebird1.safenet http://test.bluebird2.safenet http://test.bluebird3.safenet
(they’re all the same) and see how they are tomorrow

@dirvine when you say stable do you mean ongoing or that the IDs and data survives better?

EDIT: Never mind, it is obviously a test bed for client development, so it’s stable in the sense that it will be around for a while. But it isn’t intended for community-run vaults.

So if you’d like to run vaults, community1 is the only known place at present. there’s no reason not to run vaults for community1 and launchers and demo apps for this one, which is named “client_test_2” according to its config.

9 Likes

This is great about this test, you can upload some files, see if it all works and make sure the Launcher and Demo App are ok. When we have TEST 4 you can test the same things against a P2P net and find out where the pain is.

4 Likes

The droplet network must be running a later binary than testnet3 or community1. I tried to use that launcher (0.4.5) on community1 (i.e.,by substituting the community1 config file for the default) to create an account and it says “waiting for establishing connection with SAFE”. But the old launcher (0.4.3) connects to community1 straight away.

1 Like

Embrace yourselves, dirvine is gonna break the firewall of China.

12 Likes

Rabid ferocity is cool.

Rabidly testing on the droplet network with Windows and Linux launchers. :slight_smile:
Something else I notice is that you can’t run old and new launchers on the same machine at the same time: the second one will not start.

I’ll create a few accounts and let a couple of them sit untouched overnight to see if they survive.

4 Likes

Good. Break em now while you can. :slight_smile: APIs are never easy to get right, but they are harder to change once you leave beta.

4 Likes

Here’s a thought: If this network is entirely droplets then it is no wonder it is faster for the client, as compared to a network consisting of many vaults on consumer broadband. But that means it is not entirely a real-world test for the app developers. They will still need their apps to take a beating on a “P2P” network as @polpolrene called it.

3 Likes

Great to have a stable test network up again!

www.ithanium.safenet

4 Likes

It should be, as when churn handling ack mio etc are in vaults the app devs should see no difference. Otherwise we are dong something wrong :slight_smile: What we are concerned with is that we give the app devs some peace to get to grips with an API and where they can be free of us killing things as we don in network tests. This is why we are super keen not to rock the network with additional vaults / churn etc. and let those teams get to work. We will have our separate vault tests to make sure the experience is as good or better than a a churn free network as this one will be. Anyway that has to be the target in the networking side I feel.

9 Likes

OK, it is to shut them up, while you get back to work. :laughing:

3 Likes

:smiley: No not at all, they are of supreme importance to the project really so we left then at the side of the road for a few weeks there. They should now start creating majic again and thee may be a few things coming that they will like a lot.

Us lot (you included) can trash the living daylights out of network tests though, so we can ensure rock solid small resource process create a whole new eco system for the apps to run on. Not far away now :smiley: so another test maybe 2 who know and we should be in terrific shape here.

12 Likes

I’ve been doing less trashing as I learned more how to avoid crashes.

The responsiveness on this network, client_test_2, is much better, and that has to be in part because of the vaults all being droplet-hosted.

What I driving at was that a real network with many small-bandwidth clients would be more likely to experience dropped packets and time-outs, and an app developer needs to test against that eventually, or his release product will break.

1 Like

I have to say I really appreciate your prioritizing the development over the PR/marketing aspect of this (too many cryptos sometimes seem the opposite). That said, I’m looking forward to seeing what the “PR activities” entail. There’s still way too much confusion and misinterpretation (and FUD) in the crypto space with regards to MAID, so it’ll be good to clear the air and give people a better sense of what to expect. Feel free to contact me if you need some additional boots on the ground in San Francisco (I work in media, though on the literary side). Cheers, best of luck, and thanks for the ride!

6 Likes

Thank you very much for this @dirvine and the MAIDSAFE team!

4 Likes

I get Upload failures, anyone experiencing the same?

I did earlier, but it was due to exceeding the file size limit (5mb) I think.

1 Like

In my case I have 3.X MB. I also tried with smaller sized uploads. I can´t upload anything.