6 ms·

But the only time it is buggy if you SAY it's Linux. If you say it's Windows, it works fine. That's not having the software passively become buggy, that's CHANG

• emodendroket 10y ago
  It seems entirely plausible that there are optimizations that are only applied on certain platforms or something with the idea that any platform should fall back to the slower but better-supported behavior.
  • jacquesm 10y ago
    Except that we're talking - as far as is visible to me at the moment - about server side behavior, something that need not change at all depending on the user's client.
    • eridius 10y ago
      No we're not. The bug report says 100% CPU. It's very clearly doing something client-side.
      • jacquesm 10y ago
        That is speculative. Yes, the client side is the one where the problem shows up but it's when the user agent is changed that the problem appears and the client side does not do anything with that user agent, it gets sent to the server. So something in the response from the server or some interaction with the server than causes the 100% CPU load.
        • cortesoft 10y ago
          That is exactly what the people you are responding to are saying. The client makes a request to the server, and the server either returns the new optimized client code if the user agent is one that it knows supports the new code, or it sends back the older, slower, client code for an unknown user agent. They are probably returning the older code because they KNOW it works, even if slower, while they aren't sure if the new code works or not on other machines. The author says it works fine on linux when getting the windows response, but we can't be sure he has tested every use case; maybe it fails under particular conditions, which is why they are still returning the old code. Or, perhaps, they just haven't tested it so they go with the safe thing, which is to return the older code.
          • jacquesm 10y ago
            So, you're telling me Microsoft is unable to tell the difference between code that doesn't work (the 'old' code) and code that does work (the 'new' code), and that to be safe they are returning the non-working code as a fix for potential problems that do not appear to exist? If there is anything that would be more damaging to Microsoft than calling them malicious it is probably calling them incompetent, that's their core expertise you are challenging there.
            • wernercd 10y ago
              Of course MS is unable to tell the difference between code that works and code that doesn't... on an unsupported platform. Incompetent? "MS: We support Platform X, Y, Z." ... "Mob: MS must be malicious or incompetent because they aren't testing and pushing new code to Platform 9 3/4" Unsupported. Means exactly that.
              • jacquesm 10y ago
                Well, from where I'm sitting a server should just do what it does for everybody without shipping different code to 'unsupported platforms' in order to make those platforms work worse. It either serves up the same to everybody or it contains code to support some platforms and blocks others. A 'best effort but unsupported' isn't an option, it reminds me of the SMB days. Either you support a platform outright, you send out the same code to everybody or you disable the platform completely. Other options are sufficiently gray to count as ways to coerce people into making a platform switch, and the MS rep seems to be pushing in exactly that direction. " For the best experience, we recommend the operating system listed in this article. " Right.
  • pbhjpbhj 10y ago
    Should be testable, for example what optimisations are in the FF48 code sent to Windows that breaks FF48 on Linux? MS will I'm sure provide those details to demonstrate their lack of malice. /s
    • emodendroket 10y ago
      If you're that hot and bothered about it try fetching a page with both UAs.