But conversations aren't single exchanges, and shouldn't be! I mean, real life conversations are single exchanges, but this is a limitation of the format. In fact, I claim that linearizing a thread actually results in disorganization.
The tree format -- or even such things as the point-by-point quoting and response -- allows you to do things you can't do in an linear thread, in particular, keeping on multiple points simultaneously and non-disruptively.
A linear format encourages forgetting and dropping earlier points. Or, if earlier points are kept live, the result is a jumbled mess, as distinct topics alternate with each other. If instead distinct topics are segregated to different branches of the tree, they can be kept live while still maintaining organization.
In general, I would say, the proper format for writing is not necessarily linear -- nor a tree, either. (Writing my recent paper with Josh, I recall thinking how much easier it could be if I could just write it as a DAG. Nowadays you can write things as hypertext, which helps somewhat, but you can't do that in a journal. And of course you can't expect a mail client to support a DAG. Is having multiple In-Reply-To headers even supposed to be allowed?) But a tree is still better than nothing.
I'm not sure what you mean when you say little subconversations elsewhere are hard to keep track of. The software keeps track of them for you -- it organizes them into a tree, so you can look and see which other messages they relate to. By contrast, if you force them into a linear format, they won't necessarily be near the things they most relate to; you can't collapse a tree (or a more general graph) into a line in a distance-preserving way. And note that this organization means that if the subthread is about something that isn't even relevant to you, you don't have to read it! Whereas with a linear thread, it's hard to predict this in advance; the branches are bound together so you have to read them all unless you're ignoring the entire thread.
Unrelatedly -- rethinking the custom headers thing; if it's free to be a "dumping ground", that means it should be ignored entirely, because it shouldn't offer evidence either way. Spam tries to hide its spamminess, but it necessarily has giveaways in A. the parts that contain the payload, the parts the user can see and B. anything it can't fake. These are the areas that matter. Something like custom headers should be simply irrelevant.
no subject
Date: 2012-08-02 11:55 pm (UTC)The tree format -- or even such things as the point-by-point quoting and response -- allows you to do things you can't do in an linear thread, in particular, keeping on multiple points simultaneously and non-disruptively.
A linear format encourages forgetting and dropping earlier points. Or, if earlier points are kept live, the result is a jumbled mess, as distinct topics alternate with each other. If instead distinct topics are segregated to different branches of the tree, they can be kept live while still maintaining organization.
In general, I would say, the proper format for writing is not necessarily linear -- nor a tree, either. (Writing my recent paper with Josh, I recall thinking how much easier it could be if I could just write it as a DAG. Nowadays you can write things as hypertext, which helps somewhat, but you can't do that in a journal. And of course you can't expect a mail client to support a DAG. Is having multiple In-Reply-To headers even supposed to be allowed?) But a tree is still better than nothing.
I'm not sure what you mean when you say little subconversations elsewhere are hard to keep track of. The software keeps track of them for you -- it organizes them into a tree, so you can look and see which other messages they relate to. By contrast, if you force them into a linear format, they won't necessarily be near the things they most relate to; you can't collapse a tree (or a more general graph) into a line in a distance-preserving way. And note that this organization means that if the subthread is about something that isn't even relevant to you, you don't have to read it! Whereas with a linear thread, it's hard to predict this in advance; the branches are bound together so you have to read them all unless you're ignoring the entire thread.
Unrelatedly -- rethinking the custom headers thing; if it's free to be a "dumping ground", that means it should be ignored entirely, because it shouldn't offer evidence either way. Spam tries to hide its spamminess, but it necessarily has giveaways in A. the parts that contain the payload, the parts the user can see and B. anything it can't fake. These are the areas that matter. Something like custom headers should be simply irrelevant.