I think this is a good summary of the article, but not a good summary of the problem if you look at the mailing list.
This whole thing seems to have started because Thierry found a few problems with the way registers work, and wanted to address them.
The most important flaw was that after you hit C-x r SPC (save-to-register), whatever key you hit next, you'd save the text into the register associated with that key. In particular, the universal Emacs cancel key, C-g, would not work here: instead, the text or position would be saved to a register called ^g. Similarly, if you accidentally hit "jump-to-register" or "insert-register", you couldn't cancel with C-g (or with any other key), you'd be forced to select a register and it's contents would be jumped to/ inserted (if any).
Secondly, registers can hold text, a position, or nothing. jump-to-register only works if a register holds a position, while insert-register expects a text. Emacs includes a preview of registers which are non-empty when invoking these commands, but it doesn't distinguish - it will show you a register that includes strings as an option for jump-to-register, even though it knows it won't work.
So, Thierry took the time to address all of these concerns, and to address other feedback about the code. The reviewers agreed that these are important changes even though they add some extra interaction, and that the breaking change (having to hit RET after selecting a normal registry, or having to use an extra key to save to a weird register like ^g) are worth the gains. After it was done, and compiled, it was added by the Emacs maintainers to the development branch.
The author of this article came along, asked for a switch to revert between the new and the old behavior, and was asked for a patch. They provided a patch that reverted all of the changes I mentioned before (so, no way to cancel the register commands, no way to get contextual help about which registers contain text VS position, nothing) and instead implemented an entirely different feature (confirmation on overwriting a non-empty register based on a flag). Thierry installed the author's new patch and gave this simple feedback, to which the author basically replied "you don't need all that".
Now, as more people started using the feature, the belief by Thierry and the original reviewers that the breaking change had minimal impact was proven wrong. Thierry started working on a new improvement to create a flag that keeps both all of his new work, but allows the previous workflow too (particularly, removing the extra RET, which was ultimately a side effect, not the main point). The patch missed the mark, but it is still being worked on.
Overall, it seems the process is working quite well, and it is only the author of this article that is trying to bully his way into the discussion and ignore the context, with a very anti-hacker attitude of "if it kinda works, we shouldn't change it in any way", which is very much against the spirit of Emacs.
> Thierry started working on a new improvement to create a flag that keeps both all of his new work, but allows the previous workflow too (particularly, removing the extra RET, which was ultimately a side effect, not the main point). The patch missed the mark, but it is still being worked on.
I think that if the author of this rant did not come along and rant, there would be no "but is still being worked on" done. My reading is that the maintainer (Theirry) was happy with the extra RET.
NOW it is getting worked on because of the dogged and emotional criticism by the rant's author, prior to him writing his rant.
If he didn't make such an emotionally charged and disruptive arguments and patches, the rest of us poor Emacs users would have only found out about the broken workflow (extra RET) once the bug was fully baked in, requiring waiting for the next release in order to get a fix.
Makes me wonder why they didn't decree, "no using a register called ^g, that's a special character" and add a check when specifying a register. I bet it would break fewer workflows.
Well, exceptions for C-g shouldn't have to be hardcoded everywhere, because that makes it harder to move the C-g behavior to some other key. If going this route, I'd first implement a list of "keys with C-g-like behavior" that can be given more members than just C-g, via a startup flag or some such.
Then commands could just reference that list if needed.
> The most important flaw was that after you hit C-x r SPC (save-to-register), whatever key you hit next, you'd save the text into the register associated with that key. In particular, the universal Emacs cancel key, C-g, would not work here: instead, the text or position would be saved to a register called ^g
I'm not very familiar with Emacs, so possibly stupid questions incoming.
If you start to type C-x r SPC t to save to register t, does C-g work right after the C-X and the r?
If so is that because the C-x, r, and SPC are part of the command sequence whereas the t is just an argument used by the command, and the C-g handling is done in the command sequence processing?
Emacs is pretty famous for its flexibility in letting users bind and unbind key sequences to commands. Could people who like the old behavior fairly easily effectively restore it by unbinding the handler for C-x r SPC, and then bind handlers for C-x r SPC a, C-x r SPC b, C-X r SPC c, and so on for all registers they want to use, with each of those handlers just copying to the appropriate register?
That should let them use the same keystrokes they now use, and if my guess above about C-g handling is right also make C-g work to cancel after C-x r SPC.
You mean, "luckily for the entire fucking Emacs community, one of the small number of people who pick up unreleased master commits came along and happens to be a user of registers ..."
Actually multiple people have complained, some less hyperbolically than this author. And, unless you think the people who build their own Emacs from head of master are less likely to use more obscure features of Emacs, then this feedback was always going to come before anything that actually hurts the Emacs community made it even close to a release.
Having read/skimmed the bug I see now that the issues objected to by Eschel were all raised early, back in October, by the other developer looking at the work, Michael H. He pointed out the RET thing and the issue with using control characters as a registers.
This whole thing seems to have started because Thierry found a few problems with the way registers work, and wanted to address them.
The most important flaw was that after you hit C-x r SPC (save-to-register), whatever key you hit next, you'd save the text into the register associated with that key. In particular, the universal Emacs cancel key, C-g, would not work here: instead, the text or position would be saved to a register called ^g. Similarly, if you accidentally hit "jump-to-register" or "insert-register", you couldn't cancel with C-g (or with any other key), you'd be forced to select a register and it's contents would be jumped to/ inserted (if any).
Secondly, registers can hold text, a position, or nothing. jump-to-register only works if a register holds a position, while insert-register expects a text. Emacs includes a preview of registers which are non-empty when invoking these commands, but it doesn't distinguish - it will show you a register that includes strings as an option for jump-to-register, even though it knows it won't work.
So, Thierry took the time to address all of these concerns, and to address other feedback about the code. The reviewers agreed that these are important changes even though they add some extra interaction, and that the breaking change (having to hit RET after selecting a normal registry, or having to use an extra key to save to a weird register like ^g) are worth the gains. After it was done, and compiled, it was added by the Emacs maintainers to the development branch.
The author of this article came along, asked for a switch to revert between the new and the old behavior, and was asked for a patch. They provided a patch that reverted all of the changes I mentioned before (so, no way to cancel the register commands, no way to get contextual help about which registers contain text VS position, nothing) and instead implemented an entirely different feature (confirmation on overwriting a non-empty register based on a flag). Thierry installed the author's new patch and gave this simple feedback, to which the author basically replied "you don't need all that".
Now, as more people started using the feature, the belief by Thierry and the original reviewers that the breaking change had minimal impact was proven wrong. Thierry started working on a new improvement to create a flag that keeps both all of his new work, but allows the previous workflow too (particularly, removing the extra RET, which was ultimately a side effect, not the main point). The patch missed the mark, but it is still being worked on.
Overall, it seems the process is working quite well, and it is only the author of this article that is trying to bully his way into the discussion and ignore the context, with a very anti-hacker attitude of "if it kinda works, we shouldn't change it in any way", which is very much against the spirit of Emacs.