Constructing master lists of data types hga@xxxxxx (30 Sep 2019 14:51 UTC)
Re: Constructing master lists of data types Lassi Kortela (30 Sep 2019 15:09 UTC)
Re: Constructing master lists of data types hga@xxxxxx (30 Sep 2019 18:00 UTC)
Re: Constructing master lists of data types John Cowan (02 Oct 2019 17:15 UTC)
How to store the master lists of data types hga@xxxxxx (02 Oct 2019 17:52 UTC)
Re: How to store the master lists of data types Arthur A. Gleckler (02 Oct 2019 21:10 UTC)
Re: How to store the master lists of data types Lassi Kortela (02 Oct 2019 21:31 UTC)
Re: How to store the master lists of data types hga@xxxxxx (02 Oct 2019 21:54 UTC)
Re: How to store the master lists of data types hga@xxxxxx (02 Oct 2019 21:42 UTC)
Re: How to store the master lists of data types Arthur A. Gleckler (03 Oct 2019 04:11 UTC)
Re: How to store the master lists of data types hga@xxxxxx (03 Oct 2019 12:27 UTC)
Re: How to store the master lists of data types Lassi Kortela (03 Oct 2019 14:55 UTC)
Re: How to store the master lists of data types Arthur A. Gleckler (03 Oct 2019 15:07 UTC)
Re: Constructing master lists of data types Alaric Snell-Pym (01 Oct 2019 09:11 UTC)
Re: Constructing master lists of data types John Cowan (30 Sep 2019 21:59 UTC)
Re: Constructing master lists of data types hga@xxxxxx (30 Sep 2019 22:14 UTC)
Re: Constructing master lists of data types John Cowan (01 Oct 2019 20:05 UTC)
Re: Constructing master lists of data types Alaric Snell-Pym (02 Oct 2019 16:15 UTC)
Re: Constructing master lists of data types Alaric Snell-Pym (01 Oct 2019 09:33 UTC)

Re: How to store the master lists of data types hga@xxxxxx 02 Oct 2019 21:53 UTC

> From: Lassi Kortela <xxxxxx@lassi.io>
> Date: Wednesday, October 02, 2019 4:30 PM
>
>> I'm afraid that I don't understand.  Based on the subject line, it sounds
>> like this would be for storing a registry of data types supported by the
>> proposed serialization mechanism.  Do we really need more than a simple
>> text file, or perhaps a Git repo, for that?
>
> [ Git is great. ]

I'm sorry if I didn't make this clear, I am proposing to store the
*true* master list in what I think is a suitable form for git.  A log of
transactions as text that are then applied to the database to produce a
more tractable form, where one can perform joins etc. etc.

The database can be recreated from scratch by applying all the
transactions to it, can be reverted from bad ones by applying the
new ones up to the oops after starting from a checkpoint, etc.  This is
also a great tool for testing, it's an architecture I've used before
that works very well in a one low rate writer, many reader system.

This is why I've got on my Schemepersist TODO list a solid version of
such a durable log; for this interim application we can get by with a
very simple implementation, plenty of copies because it'll be small,
and text tools to edit the end if it gets messed up by a software or
hardware crash.

- Harold