Jason J. Gullickson

Jason J. Gullickson

Usenet and JSFS

I was reading about Usenet and realized how much it has in common with JSFS .

I’m generally familiar with Usenet but I’ve never been a deliberate user of it.  In it’s heyday I was too disconnected to interact with it (it wasn’t carried by the BBS’s I could dial-in to locally), and by the time that I had regular, reliable Internet access it was the dawn of the Blog Era and I was using web-based systems to meet similar needs.

For whatever reason I started digging into the implementation details of Usenet and found some similarities to JSFS.  The use of hierarchical namespaces to organize things and the idea of federating content by simply “broadcasting” data from node-to-node in a peer group.  Also the idea that any node in the network can access data stored on another node using a forwarding mechanism, and then caching the results.

I didn’t intentionally design JSFS to overlap with Usenet, and of course the things they have in common are shared with other systems that did influence JSFS’s design, but what’s curious is how features specific to JSFS could address one of the key problems with Usenet: the storage requirements.

If think you could copy Usenet’s hierarchy directly into a JSFS instance by simply replacing “/” characters used in JSFS paths with “.” used in Usenet paths.  If you did this JSFS deduplication would kick-in and presumably reduce the physical storage requirements for the archive.  Providing support for this sort of “namespace translation” could be intrinsic in JSFS or some sort of add-on module, but it’s not a lot of work.  In fact JSFS uses a dot-oriented naming scheme to establish it’s own internal fully-qualified paths.

If implemented as a module, the module could also provide an NNTP protocol interface to the archive, essentially becoming an NNTP server that is simply back-ended to a JSFS store.  Interestingly this would result in the creation of a web-accessible archive (using the standard JSFS URL’s that would result) and when JSFS indexing is implemented, a searchable archive as well.

I don’t have a specific point to make here, just that I find it interesting. I find the idea that Usenet essentially proved that the federation mechanism I’ve been considering for JSFS could actually work, and at a scale that dwarfs any other similar network that I know of.  I also like the idea of configuring federation not only by adding nodes to a peer group, but by offering the possibility of selecting subsets of the storage by namespace, making access to content via a peer fast for preferred bits, but still possible for namespaces that are less of a priority to the node operator.

I’m also very interested in exploring “store-and-forward” architectures for JSFS and distributed computing in general.  We’ve largely abandoned these designs in light of the availability of “always-on” Internet, but we need to remember that there is a cost to making the assumption that everyone can be connected all the time.  Not just the financial cost of the connection, but the social cost of cutting-off most of the world who do not have such access to the Internet.

There is also an intellectual cost that prevents us from seeing and implementing systems that can function over slow or intermittent connections. While high-speed, fairly reliable Internet has become more common and less expensive, systems that can function over slower and less reliable connections have significantly more freedom, and the hardware necessary to implement them is within reach for a much larger part of the world.  Even in wealthy countries, monopolistic business practices mean that in many communities there is only once choice for Internet access.  Network software architectures that can work on lower-tech network hardware create opportunities for community- oriented Internet access that can usurp these monopoly providers.


// jjg

Preposter.us | Github | Twitter | Ello | Google+ | Facebook