WTF Bites
-
I know we have a Teams thread and also a UI bites thread but I happened to be here when I noticed it.
I want to chat with a guy in my company. He's called Bartholomew Smith. So in the omni-search field of Teams I type "Smith" because that's shorter than "Bartholomew " (and yes I know I could just type the first few chars but it made sense to me that way).
Teams
helpfully shows me two matches, one for "Smithsonian Brown", the other for... "John Doe." Mr. Smith is nowhere to be seen.Typing "Bart..." does show me half a dozen people, including Bartholomew Smith, success!
My first thought was that the search only matched on first name (for
). That wouldn't explain "John Doe" but it would explain the other two. But then again, the search for "Bart" also shows "Mary Barton" i.e. someone where the match is on the last name.I know is hard but
.(fake ETA: now that I've actually started a chat with the guy, typing "Smith" lists him but! not in the same part of the search results ("top hits" vs "people"), the "top hits" still contains the two names mentioned above and nothing else)
-
@remi
probably because Smith is an obvious fake name
-
So I think that the issue is that the guy's last name is 3 letters long (e.g. "Doe"). I browsed my company's directory and found a few other people with 2 or 3 letters names and the same issue happens with them. I'm guessing when the search string is too short and is an exact match, it somehow fails.
Maybe someone at MS thought (inb4:
) that "nobody can have a name shorter than 4 letters so obviously below 4 we don't do exact-matches," and not only that made sense in their heads but it also actually made sense in their search system?
-
I'm guessing when the search string is too short and is an exact match, it somehow fails.
Does it search at all for very short search terms?
-
@remi Actually I think it goes like “three letters is still too short, it would still have too many matches, so only search the ‘likely’ ones for some definition of ‘likely’ that only makes sense in our head after five breaths of Skelde's pink #9 smoke.”
-
Does it search at all for very short search terms?
Yes, even on a single character it shows results (and even results that match that letter!
). Of course it only shows 5 or so and there are tons of match for a single letter, so the one I'm looking for isn't going to be there. But I'm not expecting anything else at this point, so yeah, it seems to be working.@remi Actually I think it goes like “three letters is still too short, it would still have too many matches, so only search the ‘likely’ ones for some definition of ‘likely’ that only makes sense in our head after five breaths of Skelde's pink #9 smoke.”
Probably, but that's not too different from the way I phrased it. What annoys me is that it only shows two results (which don't include the one I want). If it had too many matches (like when typing a single letter), and then failed to prioritise the exact matches above the partial matches, I still wouldn't get the result I need but that would be a more understandable issue. But this one... some sort of drug has to be involved.
-
@Bulb
strangely I couldn't reproduce with the 3 letter names 'Els' or 'Jan'. No issues either with 'Eddy' or 'Kris'. I mean no issues in regards to searching for his name ... lots of issues with Kris besides that.
-
@kazitor Insert a zero-width space after (or before) each slash? Granted I don't think CSS can do that, but the HTML side could.
-
@kazitor Insert a zero-width space after (or before) each slash? Granted I don't think CSS can do that, but the HTML side could.
<wbr>s should work too, having basically the same effect, but I’m not really in a position to change how the link is generated. Might be doable with some effort but ultimately none of this should be necessary.This isn’t just some URL thing; evidently Chrome is incapable of breaking at any slash whatsoever, so if there just happens to be an unwieldy/elongated pair like just there in the wrong spot, Chrome is going to shunt the whole thing to the next line because it’s just so smart.
Whereas Firefox will… well…
-
Browsing current C# questions on lead me to
Issue with COM Interop in VB6 After Migrating from .NET 4.8 to .NET 8
I am facing a problem with COM interop in VB6 after migrating from .NET 4.8 to .NET 8. This setup used to work perfectly in .NET 4.8 but is now failing in .NET 8. Here's the setup: I have a BaseCla...
Really? Combing (current) .Net8 with 25 years old VB6?
-
@BernieTheBernie things that remind you of
members is 
-
@BernieTheBernie 25 years is rookie numbers as long as computers aren't involved
-
@BernieTheBernie said in WTF Bites:
Browsing current C# questions on lead me to
Issue with COM Interop in VB6 After Migrating from .NET 4.8 to .NET 8
I am facing a problem with COM interop in VB6 after migrating from .NET 4.8 to .NET 8. This setup used to work perfectly in .NET 4.8 but is now failing in .NET 8. Here's the setup: I have a BaseCla...
Really? Combing (current) .Net8 with 25 years old VB6?
That's not the problem. The problem is that upgrading from 4.8 to 8.0 is still wrong, because there's still plenty of things in 4.8 that 8.0 does not support.
-
@Benjamin-Hall said in WTF Bites:
@HardwareGeek or just disallow string concatenation with + and force use of the perfectly fine string interpolation mechanism Swift has, which already does the right thing.
Swift is a bit like that cake that cat and dog made: to make the best programming language, just throw in all the nice or cool features and mix thoroughly (and the result will give you stomachache).
Calculating types when conversions, inference and/or generics are involved is always kinda slow:
- C++ only derives types forward, but custom conversions, function overloading and very flexible generics make it quite slow.
- Java and C# are fast, but they also don't have those custom conversions and their generics are much less flexible (in particular IIRC neither has associated types, not to mention so generic associated types).
- Go does infer in both directions, but its type conversion is very limited and until recently didn't have generics at all.
- Rust infers in both directions, but limits function overloading to trait implementations and also has very limited implicit conversions and still ends up being pretty slow.
- I have no experience with Haskell, but while that infers even globally (all the above require types to be spelled out in function signatures; Haskell does not), it also has no implicit conversions, and overloading only via type classes (Rust traits are really the same thing).
But in Swift they threw it in all in—inference both ways, function overloading, implicit conversions and flexible generics. And it's just too much for the poor compiler's sanity.
But it still points to there being a significant problem with the representation of types within the type checker. In particular, during analysis you don't have single types (except in degenerate cases) but instead conceptually sets of types. Which are rather messy here, but the principle still applies; the performance of type-set operations is a bit critical and sounds very naïve in Swift's implementation (it appears to be exponential in the number of type terms to unify, which is terrible).
-
@remi
probably because Smith is an obvious fake nameFun Fact: in 2010, according to the U.S. Census, the top 5 most common names were:
James Smith (20,944)
William Smith (20,590)
Mary Smith (18,845)
John Smith (17,938)
Robert Smith (16,307)
-
@Gern_Blaanston
Everybody is called Boomzilla Smith in the US
-
But it still points to there being a significant problem with the representation of types within the type checker. In particular, during analysis you don't have single types (except in degenerate cases) but instead conceptually sets of types. Which are rather messy here, but the principle still applies; the performance of type-set operations is a bit critical and sounds very naïve in Swift's implementation (it appears to be exponential in the number of type terms to unify, which is terrible).
When the problem is NP-hard, better representation won't help much, only restricting the size of the problem will. And I am pretty sure the problem is NP-hard (I never looked at Swift, but in Rust at least one subproblem is known to be, and Swift is more general).
-
strangely I couldn't reproduce with the 3 letter names 'Els' or 'Jan'.
The more I tried, the less it made sense. Some 2 or 3 letters name didn't show up at all in matches. But then some did appear, though usually in the "people" section rather than the "first hits" section that comes first. Also Teams remembers some of what you typed before so if you try the same name twice, the second time it may return different search results. Or not, depending on... something other cosmic rays, I assume?
On that topic, the worst case in one of my experiments was trying to find someone named "John Doe" because searching based on last name failed for the reasons mentioned
and searching on first name worked but "John" is so common that 5 other Johns showed up and not the one I was trying to find! So the only way to get that name to show up was to type the full name "John Doe" even though the "Doe" part was pretty unique by itself.
-
The more I tried, the less it made sense.
It's Teams. What else did you expect?
-
-
SAP
-
-
Teams
helpfully shows me two matches, one for "Smithsonian Brown", the other for... "John Doe." Mr. Smith is nowhere to be seen.I remember having a similar issue with a similarly short-named person I worked with. But I don't remember whether we figured out what was going on. Might it have something to do with their email addresses?
There might also be a customization that the Teams administrator can set for the minimum match length...?
-
Might it have something to do with their email addresses?
I don't think so, we're all in the same company and all have the same email address format (
first.last@company.com). Though if youstare long enough into the abyssdig deep enough, maybe having too-short a name in their email address somehow (frantic hand waving) matters...? But I doubt it.Also we just (literally
109 days ago) changed our company name and thus are in the middle of changing all of our domain names etc. to the new one, but AFAICS the company directory still uses the old one (and of course the old email addresses still work, and will keep doing so for some years but that's not the point here). But while that is an interesting coincidence, I don't think it's anything more than a coincidence. The search issue probably always existed and I just happened to notice it today. Still, if you want to build far-fetched theories as to why that might matter, be my guest. This is unlikely to be worse than the (unknown) actual reason.There might also be a customization that the Teams administrator can set for the minimum match length...?
Maybe, but I don't see why they would have bothered with it. Then again, I don't see why they exist in the first place (i.e. why do we even have Teams?), so your guess is as good as mine here.
-
Something happened and now Outlook Web App decided to mark all emails from external people with an
Externallabel. Which wouldn't bother me, if it didn't also put a fricking big red banner at the top of the email The sender billg@fuckyou.com is from outside of your organization. Block sender
Fuuuuuck right off.
I mean, I'd love to block all outside emails. Wait, why half-ass it? Block all emails, period. I just don't think that'd be appreciated by my boss.
-
i.e. I don't know if it's a config change on our side or on the MS side
It's a per-tenant config option for Exchange. We have the same thing enabled and it is dumb as hell.
-
@dkf we don't have the banner in Outlook, we have the banner that gets put in the body of the email at the top. It's fairly recent for us, and I assume like a lot of things added for "security" it probably doesn't actually achieve much.
-
@loopback0 We have that one too. It persists into into replies. It's dumber as hell(er) or something.
-
i.e. I don't know if it's a config change on our side or on the MS side
It's a per-tenant config option for Exchange. We have the same thing enabled and it is dumb as hell.
It is the less dumb alternative to the usual
Security solution of directly monkeying with the body of the email to put giant scary yellow boxes at the top that then become part of the reply chain to make you the laughingstock of all of your external correspondants.
-
-
@loopback0 said in WTF Bites:
@dkf we don't have the banner in Outlook, we have the banner that gets put in the body of the email at the top. It's fairly recent for us, and I assume like a lot of things added for "security" it probably doesn't actually achieve much.
I love getting email replies with that banner added to their copy of "my" reply, making it appear as if I put it there before sending the message.
-
Something happened and now Outlook Web App decided to mark all emails from external people with an
Externallabel. Which wouldn't bother me, if it didn't also put a fricking big red banner at the top of the email The sender billg@fuckyou.com is from outside of your organization. Block sender
Fuuuuuck right off.
I mean, I'd love to block all outside emails. Wait, why half-ass it? Block all emails, period. I just don't think that'd be appreciated by my boss.It's even funnier when you reply to the sender, and that warning banner becomes part of the quoted text...
How do I know?
Received such replies on my job applications...
Edit: when continuing to read the thread, I see
...
-
i.e. I don't know if it's a config change on our side or on the MS side
It's a per-tenant config option for Exchange. We have the same thing enabled and it is dumb as hell.
It is the less dumb alternative to the usual
Security solution of directly monkeying with the body of the email to put giant scary yellow boxes at the top that then become part of the reply chain to make you the laughingstock of all of your external correspondants.That and the preview in both the notification and the message list is "CAUTION this email originated from outside the business...etc"
-
It persists into into replies.
You can manually remove it, right? So what's the problem?
-
@BernieTheBernie An example (though in moon language, but anyways...)
-
@BernieTheBernie Twickling your softwareen sounds like a bad idea.
-
@HardwareGeek said in WTF Bites:
@BernieTheBernie de-
twickling your softwareen sounds like a bad idea.The ent-prefix does in this case not refer to tree people.
-
Status: I have in my possession a little box that has a light that blinks.
It apparently costs $200.
This thing better give me 100mbps throughput at that price!
-
@Tsaukpaetra said in WTF Bites:
100mbps
100 milli-bits per second? If it works at all (it's in your possession, so that's far from given), it should be able to support that throughput without difficulty. Even Milwaukee PC can do that, most of the time.
-
@loopback0
those are just check marks on someone's compliance list
-
-
-
Exhibit 1 of why type errors should display the fully-qualified name(s) of the type(s) involved.
-
Exhibit 1 of why type errors should display the fully-qualified name(s) of the type(s) involved.
Nope, this is JavaScript. No namespaces here.
instanceofdoes have some peculiarities around realms though. Maybe related.
-
I'm trying to hook up a chess engine to
.There's a primitive protocol called UCI that seems ubiquitous among chess engines. Awesome.
The protocol states that book openings are the responsibility of the GUI not the engine.
How is that GUI functionality?Ideally I don't want to have to program in the rules of chess at all, just handle board states and let the external engine do the thinking. It seems that's a pipedream, however. Moves are communicated with the source square and destination square, which I can just blindly copy one square to the other, but this fails for castling and en passant.
I guess I'll have to program chess rules after all if I want this thing to work.
-
I'm trying to hook up a chess engine to
.There's a primitive protocol called UCI that seems ubiquitous among chess engines. Awesome.
The protocol states that book openings are the responsibility of the GUI not the engine.
How is that GUI functionality?Ideally I don't want to have to program in the rules of chess at all, just handle board states and let the external engine do the thinking. It seems that's a pipedream, however. Moves are communicated with the source square and destination square, which I can just blindly copy one square to the other, but this fails for castling and en passant.
I guess I'll have to program chess rules after all if I want this thing to work.
It always comes back to Googling en passant.
-
I guess I'll have to program chess rules after all if I want this thing to work.
Well, no. You can just let folks make whatever moves they want and it will be just fine!
-
-
-
The protocol states that book openings are the responsibility of the GUI not the engine.
How is that GUI functionality? Moves are communicated with the source square and destination square, which I can just blindly copy one square to the other, but this fails for castling and en passant.
What do you mean by that? Source / destination still uniquely define the move in these cases.
Gossen Metrawatt Z501J IrDA-to-USB Interface Converter
