• Hanzito (unregistered)

    Who publishes an API, has a contract. For life. Written in blood.

  • Zatapatique (unregistered)

    Fristz!

  • WTFGuy (unregistered)

    Your API is everything you expose, nit just what's documented. Someone somewhere will take a dependency on all of it, including your quirks and bugs. It was ever thus, but the advent of the web means even "internal" code has a lot of publicly facing utput and public consumers.

  • (nodebb)

    Wrapping success responses like this which are returned over HTTP, while unfortunately quite common, is the real WTF as it doesn't solve any real problems.

    Wrapping failures in a standardised result format is a good idea, but using your own "made up" format is a WTF, you should instead be returning an RFC7807 compliant Problem Details for HTTP APIs result.

  • DeeKay (unregistered)

    Frist comment not held for moderation?

  • BWill (unregistered)

    Frist comment not held for moderation. I hope.

  • (nodebb)

    Comment held for moderation.

  • RLB (unregistered)

    I don't know whether that stereotype still exists, but yes, many of my colleagues couldn't spell to save their lives. However, it's not just programmers, it's the sales managers as well. Sometimes I think I'm the only exception.

  • Cysio (unregistered)

    I'm a fan of returning 204 if all I want to return is success

  • Roby McAndrew (unregistered)

    Ah ! The sweet spell of Sucess !

  • Smithers (unregistered)

    Next they can have the intern clear up all the misspellings in their code that say Referer:. I'm sure that won't cause any issues at all.

    Also, relevant XKCD: 1172 "Workflow"

  • (nodebb) in reply to WTFGuy

    Who publishes an API, has a contract. For life. Written in blood.

    Your API is everything you expose, nit [sic] just what's documented. Someone somewhere will take a dependency on all of it, including your quirks and bugs. It was ever thus, but the advent of the web means even "internal" code has a lot of publicly facing utput [sic] and public consumers.

    So true... Including a descriptive name together with a numeric code sounds like a good idea only until people start parsing the string. Or when people parse a browser version number as a string. Which probably just shows, that even adding a new numeric code will cause breakage, as long as it has any sequence of digits in common with an existing code...

    But yes, anything, once publically available, can become a feature people rely on. Another favorite example:

    Microsoft Excel has always had a deliberate leap year bug, which falsely treats 29 February 1900 as an actual date, to ensure backward compatibility with Lotus 1-2-3. [source: microsoft.com]

  • (nodebb)

    This is why HTTP 204 No Content exist. Indicate success but no with response content needed.

  • ZZartin (unregistered) in reply to The_MAZZTer

    This is why HTTP 204 No Content exist. Indicate success but no with response content needed.

    Except in the real world this is usually not useful as there's almost always more information needed that can't be captured in a single response code regardless of a failure or success.

  • (nodebb) in reply to R3D3

    Spolsky has an entire blog entry about that deliberate Excel date bug. https://www.joelonsoftware.com/2006/06/16/my-first-billg-review/

  • Duke of New York (unregistered)

    so the third case just hangs on an error, right?

  • Zadkiel (unregistered) in reply to Smithers

    XKCD 349, surely!

  • (nodebb) in reply to Smithers

    Next they can have the intern clear up all the misspellings in their code that say Referer:.

    Or (far, far) worse, Refferer:...

  • Splatmandeux (unregistered)

    I had to maintain some software that needed to check the Windows registry for the existence of another 3rd party program. The registry key was "Intallation Path". So... sigh... fine. Apparently someone filed an issue with the 3rd party developer and they fixed the typo in a later release. Great, now I have to search for two registry keys because both versions are widely used in the wild. Yeah, first world problems...

  • Officer Johnny Holzkopf (unregistered) in reply to colejohnson66

    Czommentz held forz sucessfull modzerasciontz. See refferer errer codez 404½+1i².

  • Randal L. Schwartz (github)

    Reminds me of the HTTP "Referer" header, spelled incorrectly by the early web server authors, but subsequently baked in to the protocol as the permanent spelling.

  • Stuart (unregistered)

    Reminds me of a guy who wanted to fix a typo in a string in a binary framework. I pointed him at a hex editor, and warned him that the string absolutely had to remain the same length or shorter, lest Bad Things Happen.

    The typo fix meant increasing the string length by one character (changing "StarTime" to "StartTime"). He ignored my warning and made the change.

    Sure enough, the "fix" didn't work. He then said that the string after that one was important else he'd steal a character from that, which would have led to... interesting results.

    The quality of education in IT professionals these days does not seem to be particularly high...

  • Your Name (unregistered) in reply to Jonathan Lydall

    The real WTF is misusing the outer protocol (HTTP) codes for content response. E.g. 404 means, the URL you requested does not exist. It should NOT mean, that the page exists, but your database has no entries for a requested items.

    What's next, using an process exit code as the size of a file you tried to open?

  • löchlein deluxe (unregistered)

    Ah yes, that reminds me of the time when our monitoring fell over because the output of yum suddenly had a new (grammatically valid) "," in its output.

  • LZ79LRU (unregistered) in reply to Stuart

    That's what you get when your developers are educated from the middle of the skill tree as opposed to the bottom. People like me who started with C or Assembly would newer make such a mistake because we understand what is going on under the hood. But if you have newer touched anything more down to the hardware than Java or Python than you really can't be expected to know or even care about things like that until you suddenly have to. And than you call one of us to fix what you can't.

  • Gearhead (unregistered) in reply to cellocgw

    Use 1904 date system: Changes the starting date from which all dates are calculated from January 1, 1900, to January 2, 1904.

  • (nodebb) in reply to LZ79LRU

    But if you have newer touched anything more down to the hardware than Java or Python than you really can't be expected to know or even care about things like that until you suddenly have to. And than you call one of us to fix what you can't.

    Out CS curriculum:

    • Java
    • Java
    • Java
    • One 2 hour lecture + 2 hour exercise for Assembler, C, and C++. Some people I knew outright skipped the assembler part ("enough to pass") and couldn't get a grip of pointers.

    Addendum 2023-09-28 05:07: Your markdown doesn seem to support the regular

    - item1
- item2

    syntax for

    • lists

    Addendum 2023-09-28 05:08: Or rather the style-sheet? The previous edit contained

    syntax for <ul><li>lists</li></ul>
  • Andrew (unregistered)

    How many English words have a different American spelling from the standard British Commonwealth? That alone can make it hard to be sure an API text string matches what people expect.

  • Maia Everett (github)

    That's a case of Hyrum's Law if I've ever seen one.

    https://www.hyrumslaw.com/

Leave a comment on “Succesful Deployment”

Log In or post as a guest

Replying to comment #:

Log Out

« Return to Article