2023-02-21 08:04:18 KipIngram: I would say most data lends itself to line-by-line, but there are always annoying cases you encounter with things like sed that don't work properly 2023-02-21 08:04:35 But going to non-line-based system is throwing the baby out with the bathwater IMO 2023-02-21 08:06:30 Sometimes it's probably better to figure out if you can translate your non-line-based data into line-based data before processing with things like sed 2023-02-21 08:07:34 Like for instance, did you know all diff tools can do binary diffs? If you convert them into a hex byte stream with one column! 2023-02-21 08:07:45 with xxd -c1 or similar 2023-02-21 08:07:57 *Should* you though? 2023-02-21 12:18:39 I think I agree with you. Not good to let the corner cases dictate the main path. 2023-02-21 12:22:14 It's good to keep in mind that by his own admission the tools Pike wrote around this time were "optimized for programmers." C doesn't care a lot about "lines," so it's not surprising that a C programmer's editor might profit from not being line-oriented. 2023-02-21 12:24:26 there might be a memory/reasoning split here; folks with better memory might want simpler tools they don't have to reason about; someone who is better at reasoning may want a complicated thing to reason through? 2023-02-21 12:31:33 Sure - I definitely think there's a "user/power user" curve associated with such things. Features that helped a beginning user "muddle through" might just crimp the productivity of a power user. I don't object to those helpful "modes" being there, but I do hate to see the power-user mode neglected. 2023-02-21 12:31:51 But power users are usually a small subset of the audience, so that neglect definitely happens. 2023-02-21 12:32:33 I think one time that tends to happen is when the GUI is "directly written" instead of being bolted atop a CLI interface. 2023-02-21 12:32:55 I.e., CLI gets added as an afterthought. 2023-02-21 12:34:48 Not thaT I've done this a bunch of times or anything, but I can't help thinking that the right pathway through that is 1) API first, 2) CLI second (just console access to the API) and then 3) GUI last (graphical access to the API). 2023-02-21 12:35:17 Then by definition you can do anything via the API, because that's the only way anything gets done period, in the end. 2023-02-21 12:59:11 that is a pretty goid application architecture 2023-02-21 13:00:52 Seems logical to me. 2023-02-21 13:01:12 But, it pospones demonstrable GUI activity to show off in scrums. 2023-02-21 13:01:30 If someone "prioritizes" the GUI over the CLI and API, well,... there you go. 2023-02-21 13:01:42 Tail wagging the dog. 2023-02-21 13:03:08 Actually you wouldn't absolutely have to do the CLI before the GUI, so long as they're both based on a complete API. 2023-02-21 13:06:01 So, the test harness I've done up the last couple of months at work - that was designed to roll through a series of test steps, allowing me to specify any traffic mix I want on each step. But... any ONE traffic mix. 2023-02-21 13:06:26 No sooner did I have it working than our big shot Fellow guy came up with a test that needed more than one traffic stream on some steps. 2023-02-21 13:06:34 So I had to write a custom test for that. 2023-02-21 13:06:59 But the last couple of days I've adjusted the system so that it now is capable of doing as many mixed streams as I want on each step. 2023-02-21 13:07:29 He also asked for a test that does trims, and the driver stack that I use (instead of the standard Linux nvme driver) won't do trims. 2023-02-21 13:07:45 So I'm also almost to the point of allowing each step to choose which driver stack to use. 2023-02-21 13:09:12 When I use the preferred stack, I call out PCI addresses explicitly, like 0000:95:00.0. But for using the Linux driver I need something like /dev/nvme0n1. Turns out you can look in /sys/bus/pci/devices//nvme pseudofile, and it gives you that information. 2023-02-21 13:10:21 So I'm not quite there on that, but it looks easy from here. 2023-02-21 13:11:14 And finally, once in a while the fio test tool "hangs." Some circumstance that I haven't tracked down yet will cause the special driver stack to unbind itself on a drive, and when that happens the fio test process just never finishes. 2023-02-21 13:11:39 If I reload the driver and kill that fio process, then my overall job just moves right on forward to the next step no problem. 2023-02-21 13:13:07 So, I think last (hopefully) tweak is to wrap the timeout command around the fio call, and if for any reason it times out instead of finishing, then I repeat that call, including making sure the driver I want is loaded. 2023-02-21 13:14:22 Sometimes having a test step not go through wouldn't be so bad - I just might miss the data for one point on a curve or something. 2023-02-21 13:14:38 But sometimes it's a critical step - necessary for getting the drive into the state that later steps need. 2023-02-21 13:14:55 So I think I need for it to operate in a way that makes sure steps are actually accomplished. 2023-02-21 13:15:24 Or maybe I write in a flag that lets me indicate whether a step is critical or not. 2023-02-21 13:16:27 Not all steps have explicit times - some of them just accomplish a certain task, like filling the drive to a particular point. But I can still put a time in those in a way that will look like a comment to fio, but that I can still grep out and use for the timeout call. 2023-02-21 21:42:06 You know, here's a phenomena that just amuses me to know end. I read this book series, The Dresden Files. It has all manner of supernatural beings in it, up to and including archangels. A perennial topic of conversation on the reddit community is "relative power levels.' I don't really know why it fascinates folks so much, but it does. Here is an excerpt from a comment someone just made to an earlier 2023-02-21 21:42:08 post: 2023-02-21 21:42:12 "This would be in keeping with the rough power levels of beings in the Dresdenverse. Angels (just regular, not archangels) are definitely an order of magnitude more powerful at least than even a Walker of the Outside." 2023-02-21 21:42:32 The thing is, NOWHERE ARE WE TOLD THAT, and nowhere is it implied are quantified in any way. 2023-02-21 21:42:55 But so many people seem to be unable to import a bunch of Christian baggage with them to the reading of the series. 2023-02-21 21:43:16 It's like they're thinking, "Oh, well - ANGELS. They've got to be more powerful that EVERYONE else. 2023-02-21 21:43:26 But the truth is there's just no data on this in the series. 2023-02-21 21:43:37 Or at least EXTREMELY limited data. 2023-02-21 21:43:57 I just don't understan where this certainty comes from, and I've read these damn things seven times at this point. 2023-02-21 21:44:11 ... and in DragonBall Z to save on animation costs the characters would spend three episodes powering up ... 2023-02-21 22:01:41 What's particularly ironic is that, while there aren't very many "outright comparisons," in one case there is - there's an overt statement (made by a character who's supposedly incapable of lying) that the Fae Queens "have power to rival the archangels and minor gods." Very overt and outright. But there is a crowd of people who just will not accept that - they will find some way to negate that bit of text 2023-02-21 22:01:42 or somehow declare it incorrect. 2023-02-21 22:02:28 And another line, ten books later, where one of those queens is cited as "basically a peer" of an archanel - same deal. There the words are, but these folks just toss up a mental barrier. 2023-02-21 22:03:05 I think it's an interesting cultural quirk. 2023-02-21 22:03:33 possibly the fantasy equivalent of those space vessels moving at the speed of plot 2023-02-21 22:03:38 I think folks, without even realizing it, important a bunch of idealogical baggage with them when they sit down to read the books. 2023-02-21 22:04:30 I should have just kept my mouth shut, but before I knew I'd replied to the above comment asking for text citations. 2023-02-21 22:05:20 "knew it"