|17 Mar 2019|
|22:37:21||#'no-defun-allowed||who the fuck were those people anyway|
|23:22:28||Gnuxie 💜🐝||spam bots|
|23:22:51||#'no-defun-allowed||i don't think they should have been able to join, but i'm not good at IRC modes|
|18 Mar 2019|
|13:08:48||mooch left the room.|
|13:19:35||mooch joined the room.|
|19 Mar 2019|
|02:57:04||#'no-defun-allowed||okay, desktop works again, so it's net🅱arm time|
|09:34:59||Gnuxie 💜🐝||That was a very long time|
|09:44:44||#'no-defun-allowed||i've been up here for five weeks, and there's RATS, nibbling at my conses|
|10:26:11||#'no-defun-allowed||So, jokes aside, I will probably write a description of the naming web model soon, and I'll finish up my new format reader and printer.|
The idea of the naming web is that we can produce human-friendly names for objects and tags for groups of objects using a web-of-trust-like graph, and we can also reduce several ordering methods into votes on these names. For example, upvoting an object could add it to your personal map, and downvoting could replace it with a "null" object on your map.
When you haven't decided, we can find new objects by looking through your friends' maps, their friends' maps, and so on. If there is a conflict, you can use a weighted vote that favours closest friends, and pick the most voted object to fit a name.
|10:34:19||#'no-defun-allowed||I also believe with collections, this eliminates a lot of moderator fluff. Going back to raddle's f/meta, a decent amount still is about what names mean.|
|20 Mar 2019|
|08:26:00||mooch2 joined the room.|
|08:28:52||mooch left the room.|
|22 Mar 2019|
|08:03:44||#'no-defun-allowed||https://pastebin.com/xrMfZhzS well, the parser, object system, and renderer are done|
|08:09:54||#'no-defun-allowed||(this is a round trip, going from the schema schema, to a serialisable object, to a block, back to a parsed vague-object, then an object, then back to the schema schema)|
|09:08:30||#'no-defun-allowed||alternatively: https://pastebin.com/7HxB0mFk has nicer printed representations since jackdaniel taught me about pprint-newline magic|
|09:11:14||#'no-defun-allowed|| (just do |
|23 Mar 2019|
|09:34:15||#'no-defun-allowed||okay, i'm making some tests, getting there|
|25 Mar 2019|
|01:54:09||mooch joined the room.|
|07:14:11||@thufir:lain.haus left the room.|
|09:58:23||#'no-defun-allowed||So, next I just have to implement a scheduler for interfacing Netfarm's dependency resolution needs with cl-decentralise's out of order loading, which should be pretty simple.|
|10:01:28||#'no-defun-allowed|| I'm not sure how to handle channels too well, but I figure that reserving channels for when user objects are created and when a user adds an object to their personal environment.|
I think I will make the environment part of the user structure, too.
In reply to @theemacsshibe:matrix.orgThe idea will be to store objects with missing dependencies, such as users and schemas into a finite-size table. We can then eliminate dependencies from the table by receiving them over the wire, or give up if there are newer objects that may be more likely to succeed at the time.
|10:16:53||#'no-defun-allowed||When I understand how to use channels effectively, I think Netfarm will be a good minimal model for distributed object storage, much like Scheme is a good minimal model for functional language theory.|
|10:22:29||#'no-defun-allowed||The only problem is that the trait system isn't perfectly reliable. A node may tell the truth, or it may tell part of the truth, which I can only detect by asking more nodes. I think Bitcoin SPV wallets have a similar problem though.|
Yes, it is an issue.
|10:26:59||#'no-defun-allowed||In Netfarm, this is problematic because we can't map trivially from one object to all the objects that add computed values to it through traits, as that would change the hash. We need a somewhat trustworthy source or all of the network context to be sure we have collected all the effecting objects.|