SAFE Browser v0.7.0 was released earlier this week (with binaries both for the mock and actual network). This is the first update to SAFE Browser since the beginning of Alpha 2 and it includes a bunch of fixes/updates.
We also have a new Android developer, called Janmejoy, starting next week based in India. This reflects our ongoing commitment to grow the team and adding more resource to our mobile team. We continue to recruit for a number of roles – Software Testing and Release Manager, UI Designer and a Software Support Engineer – to be based at our HQ and our new marketer @SarahPentland has been working with @victoria to increase the visibility of these opportunities while Victoria has been headhunting suitable candidates.
The other exciting news is that we have decided to build a fully customised SAFE Browser which will cater very specifically to SAFE functionality. Some very high-level details and links are below, however, the first step is to work toward a Proof of Concept (PoC) which will check our assumptions and further define the best path forward. Further positive news from the front-end team as they make really good progress on the Java bindings generator, and their tests into providing backward compatibility between networks suggest we may not need to shut down vaults to update SAFE APIs, although as you will see there is still some work to be done in this area.
SAFE Authenticator & API
@bochaco is updating safe_app_nodejs to the changes in the master branch of safe_client_libs. The DOM API will be updated to use the latest safe_app_nodejs after the Node.js integration is tested. The changes are not just internal, it affects the API and thus the test cases must be updated.
@shankar is working on the last phase of refactoring the web hosting example. Once that is out of the way, he is planning to start working on the developer website by grouping all documents and examples under one roof. The prototype of the dev website will be made available in a couple of weeks.
@hunterlester is focusing on advancing the CI systems to get the tests running. The browser CI is also being improved for auto-publish to GitHub releases on new tags.
We are also making some progress with the mobile platforms. @rachit has set up a sample using Gradle and we will be starting to integrate with the JNI bindings next week. On the other hand, the C# API internals are being refactored and a test framework is being integrated. The remaining APIs that were not integrated during Alpha 2 will be integrated in the following weeks.
@srini is working on an editable comments plugin, to demonstrate how versioning can be implemented using the APIs.
After last week’s Dev Forum post on Browser development and some helpful community responses there, we’ve decided to start in the direction of a fully custom SAFE Browser. So far we were using a fork of the Beaker Browser.
While browsers like Firefox and Brave were initially enticing, the complexity of using these would involve either:
- A lot of workarounds.
- Diving into core code to enable basic functionality.
Neither of which are particularly desirable.
Furthermore, when we get to look at more advanced functionality (streaming/SAFE-specific functionality) things start to look even more complex.
As such we’re planning to start working on a fresh Electron-based browser. This will let us use the knowledge we’ve gained so far, but work in a much cleaner fashion and focus on what
SAFE: needs. So we’re planning to release a basic
SAFE only browser as a first step in the next couple of weeks. We consider this as a POC for the custom browser. Based on the stability and feedback of the custom browser, we will be deciding on whether to continue on this route or to rethink on other options.
All going well there, we’ll then look to reintegrate the authenticator functionality before planning if/what other features we want to see in the new browser.
SAFE Client Libs
We are continuing to think about how to maintain backward compatibility between the networks. Our recent tests show very promising results: the recent changes in the master branch are compatible with the data currently stored on the Alpha 2 network, which means that we won’t need to shut down vaults to upgrade our APIs. @adam is currently working on serialisation tests to confirm this and to make sure that this compatibility is maintained in the future. However, as promising as it is, we’ve yet to find a way to make core data structures future-proof, like in the case when we want to introduce multi-sig to Mutable Data. We’re looking for various options to make it unnecessary to shut down Vaults even in the case when we need to introduce new capabilities for our data structures. As mentioned in the previous update, we plan to write and publish a comprehensive RFC document on this topic at some point in the future.
We expanded our
safe_app API by implementing new functions to the
crypto module (MAID-2412). We started by adding a new object cache handle for secret signing keys named
SignSecKeyHandle and replacing the existing
SignPubKeyHandle. We provided functions for dealing with the new secret key handle, such as
sec_sign_key_get, which we modeled after the equivalent functions for public key handles. We also added the function
sign_generate_key_pair which generates a pair of such keys and returns the handles. Finally, we added two new functions
sign can sign data using either an app’s own secret key or any arbitrary secret key handle, while
verify takes some signed data and returns either the original plaintext or an error if it could not be verified. Overall, the new API functions should feel familiar as their usage is consistent with existing functions.
In addition to the above, we have merged a fix for app reauthorisation using 2 PUTs and an API change replacing value structs with pointers.
Lastly, we’re having good progress with the Java language bindings generator and we should see first tangible results in the next week.
Routing, Vault & Crust
The Routing design discussions have continued this week with the focus being on the relative pros and cons of the two remaining proposals, namely vanilla Data Chains and the Section Graph variant. They’re intentionally very similar options, but have a couple of fundamental differences. We’re now leaning towards Data Chains, but still have to flesh out some more details before we embark on the real low-level design task. The areas we’re looking into include penalties for misbehaving peers (including how to detect them, how this could be exploited, the severity of penalties), peer ageing and handling mass peer loss across the network.
No major changes in Crust this week but lots of minor ones. We’ve been hard at work improving documentation and expanding our test suite, which has enabled us to catch and fix a couple of bugs. Thanks to that, connections should no longer hang when they’re being closed, and config-file monitoring should now work reliably. In the next few days, Crust should get to the point where we’ll be ready to start integrating our NAT traversal and uTP libraries.