Hacker News Reader

Comment by mhluongo | Hacker News Reader

They can't be throttling based on user-agent... right? I thought this was the New Microsoft. Any other hypotheses?

It's not throttling based on user-agent. It's just a bug that happens when the user-agent parser falls off a cliff when it exhausts the values that it knows about.

A bug that Microsoft won't fix, because Linux is not supported.

</charitable>

But user-agent parsing should be blazing fast, a one-liner regexp or similar, isn't it?

That was sarcasm

I don't think so, I think the person was utilising Hanlons razor[0]:

Never attribute to malice that which is adequately explained by stupidity

[0] https://en.wikipedia.org/wiki/Hanlon's_razor

Gotta create a good name for the "never attribute to stupidity something that makes the perpetrator rich" maxim.

Yes, I intended it as sarcasm.

The user-agent parsing is fast, it's the server-side rendering logic specific to Linux user-agents that is somehow terribly slow.

Anecdotal experience... but when I use Chrome and switch user agents to a random one that probably wouldn't be too common (in my case, Mozilla/5.0 (X11; OpenBSD i386) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/36.0.1985.125 Safari/537.36), in Chrome Network tools it seems that there were several Javascript pages that were being loaded through a "StaticLoad.aspx" (namely WordEditor.js and Hermes.js scripts), that simply hung while downloading.

Changing the version didn't seem to matter (I couldn't find the latest one for any Linux style OS easily, but I tried "Mozilla/5.0 (X11; OpenBSD i386) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/56.0.2924.87 Safari/537.36" with the same results). However, Android type user agent strings like "Mozilla/5.0 (Linux; Android 4.0.4; Galaxy Nexus Build/IMM76B) AppleWebKit/535.19 (KHTML, like Gecko) Chrome/18.0.1025.133 Mobile Safari/535.19" load fine.

Not knowing the code I have no idea what this could mean, but I suspect you're 100% right here, there's some server side rendering that does different things based on user agents, and it's having trouble on the "other UA" category for some reason.

It seems something more like:

    case @useragent
      when 'Windows'
        send file fast
      when 'OSX'
        send file fast
      else
        sleep 3600

More likely something like:

resolve_renderer_module(user_agent.os)

where the osx/Windows ones are already loaded in RAM, but it has to try (and fail) to load the linux one every time somebody tries.

This is classic Hanlon's razor stuff.

Is it possible that companies, teams within companies, and/or individuals might exploit "Hanlon's Razor" to their own benefit? It's easy to say "Oh, whoops" and often there's no easy way to prove incompetence vs malice. And because people are aware of this cliche/meme, they may be more susceptible to it.

I wonder if people said "Hanlon's Razor" when they first heard about the VW emissions scandal.

NOTE: I'm just thinking out loud here. I don't have any particular bias against Microsoft.

> I wonder if people said "Hanlon's Razor" when they first heard about the VW emissions scandal.

I don't think that was possible, because the engine specifically performed [0] better during the test and worse during normal operations. So what would Hanlon's Razor be? Whoops, the testing procedure accidentally happened to make our engine perform better?

[0] With the definition of performance being emissions output, not power or torque.

yes, but the hypothesised cause is an unhandled exception when it doesn't find a UA string from its list of expected UA strings, and that case might just hang until a timeout of some kind.

I have no direct understanding but I have a hypothesis;

- different generated html / js depending on user agent (UA)

- bug fixes are rolled out on a UA by UA basis

- bug fixes have been rolled out for Windows and Mac (supported)

  - but not for the "generic" HTML / JS output

If this is true it will mean that anything other than the supported platforms will fall further behind as time goes by.

That sounds highly likely. Intentional throttling doesn't help Microsoft in this case. I don't think they view Linux desktops (aside from Chromebooks) as a serious threat. Why wouldn't they want more OneDrive users?

This reminds me of the article about the OneDrive team being allowed to introduce ads into Explorer because they don't care at all about the experience of using Windows and they just want more OneDrive users.

(Supporting documentation of the position "why wouldn't they want more OneDrive users"...)

https://news.ycombinator.com/item?id=13903701

A company has an IT director who upon review has determined that Linux workstations would be acceptable for some of his users now due to the fact he uses onedrive.

He tests this solution, finds it slow and decides to stick with Microsoft.

This is probably a happy accident but one they won't move quickly to fix.

Throttling does help if they can get away with it. Now where did I put my windows phone...

> depending on user agent (UA)

This would be very amateurish. The right way to do it is to check browser capabilities. User agents can be easily manipulated and are considered unreliable.

And not every browser bug is detectable via feature detection. I've seen plenty of render specific bugs that were only able to be targeted via user agent parsing. Not it isn't perfect... but it's often the better of the options available.

I'd really love to see what their excuse is.

They're probably doing broad bucketing for supported browsers first, instead of assuming it's supported then limiting/targeting specific versions for various fixes, or non-support.

Hey why doesn't someone just check the sourcecode? ...

That sounds exactly like Google, and makes the most sense.

If this happens to also drive people to Windows, that’s just a side bonus.

Regarding Google: Visit google.com with Firefox Mobile UA, and Chrome Mobile UA. The version presented to Firefox is the Google 2010 version, basically unchanged, with no search options or previews. The Chrome Mobile version actually works.

This is part of why Firefox Mobile now supports all -webkit- properties, webkit’s extensions of JS partially, and emulates Chrome’s UA for a great deal of websites.

I read in a comment on Reddit (lost the link at the moment), that it doesn't appear to be nefarious. Basically, they're trying to do feature detection based on UA instead of.. actual feature detection.

Google is also doing this with Universal 2nd Factor working on Chrome but not on Firefox (which supports U2F with an addon).

Can't tell if this is intentional though.

Plenty of parts of Google have done UA-sniffing all over the place. Opera, even after being Chromium based, have had plenty of problems as a result: the UA doesn't match so therefore they get an "unsupported" or "degraded" experience.

Even basic Google search seems to do some UA sniffing. I run Firefox but usually browse with IE or Chrome user agents, and clicking the results on a search won't actually navigate to them

It's just a very heavy page render due to the way asset preloading is implemented; details at [1]. Seems Linux user agents get a legacy method and Windows get a modern one.

You can see it yourself by viewing [2]. OneDrive renders several of those pages simultaneously.

[1]: https://productforums.google.com/forum/#!msg/chrome/x-B5T7tn...

[2]: https://powerpoint.officeapps.live.com/p/StaticLoad.aspx?wx=...

> They can't be throttling based on user-agent... right?

It can be, but not for the motives many would assume. I can imagine them white-listing clients that they've thoroughly tested with advanced functionality, and blocking it for others.

I'm not for a second suggesting this is a wise course of action, but I also really don't think it's a cynical attempt to make Linux a less viable platform. It's just cover-your-ass behaviour for a corporate-focused product.

User agent sniffing is usually a bad shortcut for feature detection, and little more.

There can be a feature that is only tested on Windows and so is only enabled there. Does it make sense? Not really, but that can happen. Google does this everywhere, blocking features on browsers that are not Chrome.

Well obviously they're trying to prevent the year of linux desktop.

They're still trying to get even for Marc Andreessen sabotaging NCSA Mosaic's cross-system portability stuff, in an attempt to prevent the year of the Windows Web Browser.

https://github.com/alandipert/ncsa-mosaic/blob/fd11fb7c900e8...

  Much of the cross-system portability stuff has been intentionally
  REMOVED from this version of the library by Marc A in order to
  discourage attempts to make "easy" ports of Mosaic for X to non-Unix
  platforms.  The library needs to be rewritten from the ground up; in
  the meantime, Unix is *all* we support or intend to support with
  this set of source code.

And of course for this remark:

https://ontechnologyandmedia.com/2009/07/12/the-return-of-th...

Netscape co-founder Mark Andreessen boasted that Netscape would reduce Microsoft Windows to “a set of poorly debugged device drivers.”

You can't teach an old microsoft new tricks.

I thought this was the New Microsoft. Any other hypotheses?

Yes: there is no single 'New Microsoft'. I don't even know where the term originated. Instead there are a couple of departments within MS which have been heading in a new direction (see the open-sourcing, VSCode, ...) but others seem still to be stuck in <don't know what to call it>. Mismanagement maybe?

Some parts of Microsoft are newer than others.

>Any other hypotheses?

Other than the obviousness of the new MS being worse than the old MS? Between the major issues Windows 10 has in almost all respects but not limited to shocking privacy issues, the very poor quality of updates, and the hamfisted way Win10 was forced on some users, I suspect we'll be missing the Gates/Balmer years more and more. As someone who has to roll out 100 Win10 workstations, I can't believe the shit I have to jump through to get basic funtionality and how the defaults in terms of start menu items, included apps, privacy, gpo limitations, etc is highly business inappropriate, even with professional volume licenses.

I'm not sure what the "New Microsoft" really means, but from what I've seen its something that has only added hassle to enterprise and chased more people into the loving arms of OSX. Everyone in my social circle now has a mac after toughing out on 7 and giving up on 10. Nadella's run is starting to look like a Marissa Mayer-like run to me. A welcomed change of leadership from the tech press but one that was unable to fix historic problems and also introduced their own new set of problems.

Throttling is the only thing I can think of. This is the New New Microsoft then, similar to the first.

"And sometimes there's a third, even deeper level, and that one is the same as the top surface one.

Like with pie."

Like with onions.

It is the "new" Microsoft, in terms of PR. Kind of like AT&T's "Gigabit to the press release." Tons of positive press for the announcement, but nothing real behind it.

New outside, same old inside.

Reply on news.ycombinator.com