- From: Christopher Lemmer Webber <cwebber@dustycloud.org>
- Date: Sat, 17 Aug 2019 09:35:43 -0400
- To: public-credentials@w3.org
- Cc: Serge Wroclawski <serge@wroclawski.org>
As promised, here's that rant. It's not actually all that ranty. ;) So having seen Datashards and DB's work on Secure Data Shards, and knowing that I occasionally contract for DB, you might think that maybe DB also had something to do with Datashards. That's not the case, other than that I've been trying to convince them of the ideas for the last few years (we do talk a lot, but I haven't been paid for any work at all in this area). Over time we've come closer, but our work has happened seperately, and we keep checking in and seeing that we come to many of the same conclusions for different reasons, but not all of them. So the ideas are close, but they aren't yet aligned. I hope at RWoT we can get some better convergence. There is one major difference that, in truth, barely matters (except that it kind of does): - A difference in how data is serialized. We can discuss this, it's not too important though. It's easy enough to get convergence on this issue. There are also several things that I used to argue for that we've already come to convergence on: - The need for symmetric encryption on content-addressed content (well, at least we agree it's needed for some things; I'll argue for all CAS things later) - The need to chunk content into 32kb chunks Yay, convergence! Then there are several different differences which actually matter: - Datashards not only specifies an immutable system, but also a mutable system that layers on top of the immutable data. The design is mostly taken from Tahoe-LAFS (which has done almost everything right except for making their system ~universal in a way that can be applicable on the web... in Tahoe you have to run things on specific clusters). - Datashards has two universal URI schemas that can exist *outside* of a data hub. idsc: is for immutable data, mdsc: is for mutable data. This is, I would argue, a nice feature. You can use these for everything you would normally put an https: link; this includes json-ld contexts and linking to cat photos and all sorts of things. (This is what I mean by Datashards building "universal web primitives".) - Datashards can exist *anywhere*. You can put them on a usb key, you can stick them in a key-value web store, whatever. I'm arguing that Secure Data Hubs are a good idea, except that they should just be "a common web interface to store your Datashards". That's a big feature because then you have a URI so you can refer to those same objects univerally, the same way you an refer to an http uri, etc. - I think DB isn't yet convinced that we need symmetric encryption for *all* content-addressed content. Again, I refer you to the podcast for why this is, but I think that content-addressed peer to peer systems that don't do symmetric encryption and chunking by default not only don't have a nice privacy layer by default, but they put node operators at risk. (Imagine if you worked for a volunteer postal delivery service, but instead of packages and letters being wrapped in opaque paper, they were wrapped in syran wrap. You might start to feel like you need to make moral judgements for what you're delivering, and even worse, making moral judgements may become so commonplace that eventually a law is enacted telling all postal service volunteers that they *must* examine contents before delivering them, and are likewise liable for their contents. Sometimes it's better to know less.) We disagree on the four above points somewhat, but I'll argue that it's really little to no extra cost (it may even be less) to switch the Secure Data Hubs approach to the way I'm describing (being *a* place to store Datashards as opposed to being content which can only live at that node and can't be conveniently referenced outside of it), and the gains are very great. But, RWoT is a nice place for us to work on resolving these differences, which I am convinced are really not so large, but the benefits of resolving them are very high. I look forward to working through them at RWoT. :) - Chris Christopher Lemmer Webber writes: > Since I intend to bring up both in a significant way at Rebooting Web of > Trust, it may be useful to hear that we put out two episodes on the > podcast I co-host: one about Datashards: > > https://librelounge.org/episodes/26-announcing-datashards.html > > and one about OcapPub: > > https://librelounge.org/episodes/25-managing-spam-and-hate-speech-on-the-fediverse.html > > More about Datashards: > https://github.com/WebOfTrustInfo/rwot9-prague/blob/master/topics-and-advance-readings/datashards-rationale.md > > More about OcapPub: > https://gitlab.com/spritely/ocappub/blob/master/README.org > > You may wonder how Datashards and Secure Data Hubs can fit together. > For that, I will follow up to this email with a rant. ;)
Received on Saturday, 17 August 2019 13:36:07 UTC