Notices by Garrett Wollman (wollman@mastodon.social), page 2

@EugestShirley @MsMerope @ai6yr I have about a hundred cookbooks, but I am very lazy and do little scratch cooking, so the ones I use most are baking books. (Some of them, like Joanne Chang's FLOUR and FLOUR TOO, have cooking recipes in them too.) Probably Claire Ptak's THE VIOLET BAKERY COOKBOOK, King Arthur Flour WHOLE GRAIN BAKING, and Alice Medrich's SERIOUSLY BITTER SWEET are the ones I refer to most in addition to Chang's. (I do have 4 editions of JOY as well, more for reference than use.)

@emaste You know, I'm still contemplating how I'm going to upgrade from 13 to 14 so that's pretty far off my radar at this point. I'm still waiting for pkgbase to be complete (including an etcupdate replacement and a reliable way to adopt existing systems) but I may be retired before that happens.

After supper tonight, I'm going to try upgrading three #PostgreSQL "clusters" from 14 to 15. I don't think there are any anticipated gotchas, but that's why we do these things on little personal databases before taking a wrecking ball to prod.

Still going to have to research how to upgrade a "cluster" that is running streaming replication. That'll be a hassle for sure.

The #PostgreSQL upgrade (14->15) was shockingly easy on all three servers, which makes me wonder what horrors it has in store for me when I go to do the same thing on my work database servers, which have much more complicated access controls, replication, and other fun stuff like that.

@rubenerd @paoloredaelli @dexter I still have Perl CGI scripts in production that I or colleagues wrote 25 years ago. To replace them with something "modern" would require three times the resources and at least several hundred hours of dev time (and it would be very difficult for said dev, who we do not currently have, to resist the temptations of second-system effect to completely rototill the data model and infrastructure just because it was designed in the year 1999).

Most emacs users will have their terminal configured so that ctrl-SPACE sends NUL, even though it should be ` (backtick), because set-mark-command is used so frequently and C-SPC is much easier to type than C-@. I don't remember what we used to use to set the mark when we were stuck with non-NUL-clean communications channels, back in the 1980s.

There are seven control characters that few people are aware of (at least of their being control characters) and are kind of hard to type. Three or four of them are likely to be usable for user-specific purposes without interfering with application keybindings:

^@ (= NUL, set-mark in emacs)
^^ (I use this for screen because nothing else does)
^_ (undo in emacs)
^\ (usually bound to SIGQUIT, causes core dump)
^[ (= ESC)
^] (used to break out of telnet)
^? (= DEL)

^X means chr(ord(X) xor 64)

Pleasant #FreeBSD surprise of the day: rpc.rquotad(8) works on ZFS, despite the fact that quotactl(2) is documented as only working on UFS, and indeed the fact that the quota models implemented by UFS and ZFS are quite different.

Unfortunately, it's Yet Another ONC RPC service when we would prefer for there to be fewer of them (as we are moving our NFS servers to v4-only specifically to be able to kill off portmap and all the other non-port-2049 crud).

@feld I wouldn't want to do that at boot time, only upgrade time, and I'd want to be certain that both loaders were updated before booting the upgraded system. If we had some conventions about how the ESP was identified, the install process could do it automatically, but at least all of my systems are sui generis -- I have no idea even how the stock installer sets things up. Ideally, there could be a pkgbase trigger to do it. Of course you don't want to overwrite every visible ESP willy-nilly.

@edolnx That's not the issue. The problem is making sure that the ESPs on both drives get updated at the same time. I could do that with a manual/"unlabeled" gmirror, maybe, but if they diverage? Currently what I'm doing is just dd'ing the fresh ESP onto the stale one, because that's easier and more certain than newfs/mounting each one separately and copying loader.efi in. The firmware doesn't seem to object to this; those machines still boot fine.

Reporting something positive for a change: at work, I recently retired the last six of our 13-year-old Dell R415 servers, which were replaced with a combination of new Dell R360s and 7-year-old whitebox Supermicro machines. I cloned the old machines onto the new machines' disks, made small configuration tweaks (UEFI, serial console settings, network interface configuration) and after a couple of initial hitches everything Just Worked. #FreeBSD of course.

The most annoying thing about Ansible is that it prompts for passwords before validating any input or command-line parameters.

@ripienaar @b0rk I think bash has always had `source`, which is a csh-ism that was never in the Bourne or Korn shell, and it's always been equivalent to the `.` command going back to pre-1.x days.

@ripienaar @b0rk Quoth The Standard:

If file does not contain a <slash>, the shell shall use the search path specified by PATH to find the directory containing file. Unlike normal command search, however, the file searched for by the dot utility need not be executable. If no readable file is found, a non-interactive shell shall abort; an interactive shell shall write a diagnostic message to standard error, but this condition shall not be considered a syntax error.
(SUSv7 but unchanged I think.)

@b0rk I guess it is faster, in a very technical sense involving how the tty driver and line disciplines were implemented that you probably don't want to get into -- since every raw-mode application (like readline) has to context switch back to user mode so the application can read the terminal buffer and interpret the control codes. This is also why there's also a mode where the kernel interprets controls that send signals, but the application does everything else.

@b0rk (This is what allowed Big Iron like System/370s to handle thousands of simultaneous users, whereas Ken and Dennis were not even trying to do that on their little PDP-7 and PDP-11.)

@b0rk Nope! Unix has always used "character" terminals that contained no local editing logic, just depending on the host to do everything. IBM mainframes and similar systems tended toward "block mode" terminals like the 3270, where the host would transmit an entire form to be filled out remotely, and then when the user pressed "send" it would get transmitted to the host. This was done because interrupts were slow and expensive so you didn't want to take one for every key pressed on the terminal.

Anyone else seeing ssh script kiddies trying to log in as `%split%`? I guess their script must have a bug. (This is presumably a variant of the attack where they try to log in as various components of the FQDN, but I guess they screwed it up somehow.)

Decided to buy a 52-week T-bill this month rather than the 26-week bills I have been rolling over for a couple years now. The rates recently haven't been that much worse, and the Treasury market is likely to be fucked next month because of the budget/debt ceiling stuff (the Treasury is already into "extraordinary fiscal measures").

At today's auction:
26-week 4.372%
52-week 4.237%

In fact, every maturity *except* 52-week cleared the auction at a few bps over 4.3%. #investments #uspol

What is the origin of "I hope this message finds you well"? People did not write that 30 years ago, or even 20 that I can recall. What community did it escape from, Victorian England?