Possible HTTPS bug in IE9 testers?

Support
1

Hi Guys,

Wondering if I could get a little help.

I think this is potentially a bug with HTTPS in IE9.

Neither of the following tests are for websites I work with. I do appologise for using other websites. But, I am not authorized to publicly discuss the websites I work on. I hope you will understand.

Both of the following tests are examples of what happens when I try to test the HTTPS version of my site on public IE9 WPT test instances.

[url]http://www.webpagetest.org/result/150930_DD_V4X/[/url]
[url]http://www.webpagetest.org/result/150930_TK_TR3/[/url]

Looking at the waterfalls you will see there are very long periods of inactivity. This type of test result only happens in IE9 and only with HTTPS.

These types of results have been occurring for, I think, several months

If someone could help me sort through whether or not this is a bug, I would greatly appreciate it.

Thanks

BeauFLA

2

Same to me…everthing gone…

[url]http://www.webpagetest.org/result/151005_V5_5CW/[/url]

3

@lelongbank - Just re-ran the test and it looks like it worked: http://www.webpagetest.org/result/151005_V8_Q29/

Could have been a transient problem or something else might be going on. Can you let me know if you see it again?

@BeauFLA It’s not a known issue but maybe something strange is going on with EV certs. I re-ran one of the tests and one out of the three test runs looks normal while the other two have the long 15-second delays.

Testing google.com didn’t show any issues so it has to be something different with the type of certs, the chain, validation or something like that: http://www.webpagetest.org/result/151005_8D_QSP/

I ran the test with tcpdump enabled and see DNS lookups for g2.symcb.com, g1.symcb.com and gn.symcd.com which look to be OCSP responders for Geotrust so the time gap definitely has something to do with the OCSP/CRL verification checks but the strange part is that I don’t see requests actually being made to the responders (even in the tcpdump).

4

Hi Patrick,

–Edit–

As far as one out of three tests running normally is concerned, I’ve seen this behavior as well. Makes the issue seem intermittent when it really is not. I test a block of pages on our sites weekly, the pages affected can change weekly.

–End Edit –

Thank you for looking into this issue. I’m gonna look into it a bit further on our end. If I find anything interesting, I’ll update this thread.

Thanks Again,

BeauFla

5

Hi Patrick,

I was putting some more thought into this last night. And, just for giggles I reran the test you did for Google.com.

[url=http://www.webpagetest.org/result/151008_FK_904/]http://www.webpagetest.org/result/151008_FK_904/[/url]

And then I ran the test again

[url=http://www.webpagetest.org/result/151008_ED_AT2/]http://www.webpagetest.org/result/151008_ED_AT2/[/url]

Both tests show the same delay in HTTPS.

I also ran a test for a wikipedia page - LOL, I was looking for a list of HTTPS pages to test.

Here is the result.

[url=http://www.webpagetest.org/result/151008_75_8MW/]http://www.webpagetest.org/result/151008_75_8MW/[/url]

It also shows the same type of delay.

I also loaded a good number of pages from my HTTPS sites on my IE9 VMs and never once had any extreme page load times. Granted, it was only a visual assessment. But, 15 to 19 second delays are rather difficult to miss.

I’ve come full circle on this issue. I really think this is a bug in the IE9 testers with HTTPS.

BeauFLA

6

Sorry for dragging it out so long. Not sure if the VM’s got compromised or something just broke (my bet is on compromised). I just nuked them and configured a set of new, clean IE9 VM’s and it looks to be behaving much better: http://www.webpagetest.org/result/151009_3G_3MK/

Let me know if you see any more issues.