Bugzilla

Hey sancus, not every request takes 4000 ms, but some do. For example, I do not see that problem on my machine. The tests were performed on Neil's machine. He saw it in some, but not all requests. Still, it was fairly reproducible for him. He could code and test against these timeouts.

Of course, the OSCP checks are cached for a certain amount of time, so you would only see this in a new profile. That made it harder to discover the bug.

> this kind of delay can't possibly be acceptable, normal operation and, if it is actually widespread, should be considered a bug.

I concur, and I am glad you  consider it the same. I don't know any tools or facilities that would both allow me to query from different DSL connections around the world, and also do it in the same way as Thunderbird does, with OSCP. Commandline tools normally don't bother with this, as evidenced by the good and fast 'hay' results that Neil saw on the same machine where Thunderbird requests are slow.

I can only go by symptoms, and if the server calls don't give a reliable response, that would explain a number of symptoms that I see that were inexplicable so far. But I cannot be certain.

> autoconfig latency is poor(~800-1100ms) in many regions

While 1100 ms is certainly not fast, it's still acceptable. I would worry upwards of 2000ms, or outliers above 4000ms.

Hey sancus, not every request takes 4000 ms, but some do. For example, I do not see that problem on my machine. The tests were performed on Neil's machine. He saw it in some, but not all requests. Still, it was fairly reproducible for him. He could code and test against these timeouts.

Of course, the OSCP checks are cached for a certain amount of time, so you would only see this in a new profile. That made it harder to discover the bug.

> this kind of delay can't possibly be acceptable, normal operation and, if it is actually widespread, should be considered a bug.

I concur, and I am glad you  consider it the same. I don't know any tools or facilities that would both allow me to query from different DSL connections around the world, and also do it in the same way as Thunderbird does, with OSCP. Commandline tools normally don't bother with this, as evidenced by the good and fast 'hay' results that Neil saw on the same machine where Thunderbird requests are slow.

I can only go by symptoms, and if the server calls don't give a reliable response, that would explain a number of symptoms that I see that were inexplicable so far. But I cannot be certain.

> autoconfig latency is poor(~800-1100ms) in many regions

While 1100 ms is certainly not fast, it's still acceptable. I would worry upwards of 2000ms, or outliers above 4000ms, including all redirects commulative.

Hey sancus, not every request takes 4000 ms, but some do. For example, I do not see that problem on my machine. The tests were performed on Neil's machine. He saw it in some, but not all requests. Still, it was fairly reproducible for him. He could code and test against these timeouts.

Of course, the OSCP checks are cached for a certain amount of time, so you would only see this in a new profile. That made it harder to discover the bug.

> this kind of delay can't possibly be acceptable, normal operation and, if it is actually widespread, should be considered a bug.

I concur, and I am glad you  consider it the same. I don't know any tools or facilities that would both allow me to query from different DSL connections around the world, and also do it in the same way as Thunderbird does, with OSCP. Commandline tools normally don't bother with this, as evidenced by the good and fast 'hay' results that Neil saw on the same machine where Thunderbird requests are slow.

I can only go by symptoms, and if the server calls don't give a reliable response, that would explain a number of symptoms that I see in the statistics that were inexplicable so far. But I cannot be certain.

> autoconfig latency is poor(~800-1100ms) in many regions

While 1100 ms is certainly not fast, it's still acceptable. I would worry upwards of 2000ms, or outliers above 4000ms, including all redirects commulative.

Hey sancus, not every request takes 4000 ms, but some do. For example, I do not see that problem on my machine. The tests were performed on Neil's machine. He saw it in some, but not all requests. Still, it was fairly reproducible for him. He could code and test against these timeouts.

Of course, the OSCP checks are cached for a certain amount of time, so you would only see this in a new profile. That made it harder to discover the bug.

> this kind of delay can't possibly be acceptable, normal operation and, if it is actually widespread, should be considered a bug.

I concur, and I am glad you  consider it the same.

Whether this is widespread: I don't know any tools or facilities that would both allow me to query from different DSL connections around the world, and also do it in the same way as Thunderbird does, with OSCP. Commandline tools normally don't bother with this, as evidenced by the good and fast 'hay' results that Neil saw on the same machine where Thunderbird requests are slow.

I can only go by symptoms, and if the server calls don't give a reliable response, that would explain a number of symptoms that I see in the statistics that were inexplicable so far. But I cannot be certain.

> autoconfig latency is poor(~800-1100ms) in many regions

While 1100 ms is certainly not fast, it's still acceptable. I would worry upwards of 2000ms, or outliers above 4000ms, including all redirects commulative.

Hey sancus, not every request takes 4000 ms, but some do. For example, I do not see that problem on my machine. The tests were performed on Neil's machine. He saw it in some, but not all requests. Still, it was fairly reproducible for him. He could code and test against these timeouts.

Of course, the OSCP checks are cached for a certain amount of time, so you would only see this in a new profile. That made it harder to discover the bug.

> this kind of delay can't possibly be acceptable, normal operation and, if it is actually widespread, should be considered a bug.

I concur, and I am glad you  consider it the same.

As to whether this is widespread: I don't know any tools or facilities that would both allow me to query from different DSL connections around the world, and also do it in the same way as Thunderbird does, with OSCP. Commandline tools normally don't bother with this, as evidenced by the good and fast 'hay' results that Neil saw on the same machine where Thunderbird requests are slow.

I can only go by symptoms, and if the server calls don't give a reliable response, that would explain a number of symptoms that I see in the statistics that were inexplicable so far. But I cannot be certain.

> autoconfig latency is poor(~800-1100ms) in many regions

While 1100 ms is certainly not fast, it's still acceptable. I would worry upwards of 2000ms, or outliers above 4000ms, including all redirects commulative.

Hey sancus, not every request takes 4000 ms, but some do. For example, I do not see that problem on my machine. The tests were performed on Neil's machine. He saw it in some, but not all requests. Still, it was fairly reproducible for him. He could code and test against these timeouts.

Of course, the OSCP checks are cached for a certain amount of time, so you would only see this in a new profile. That made it harder to discover the bug.

> this kind of delay can't possibly be acceptable, normal operation and, if it is actually widespread, should be considered a bug.

I concur, and I am glad you  consider it the same.

As to whether this is widespread: I don't know any tools or facilities that would both allow me to query from different DSL connections around the world, and also do it in the same way as Thunderbird does, with OSCP. Commandline tools normally don't bother with this, as evidenced by the good and fast 'hay' results that Neil saw on the same machine where Thunderbird requests are slow.

I can only go by symptoms, and if the various server calls don't give a reliable response, that would explain a number of symptoms that I see in the statistics that were inexplicable so far. But I cannot be certain.

> autoconfig latency is poor(~800-1100ms) in many regions

While 1100 ms is certainly not fast, it's still acceptable. I would worry upwards of 2000ms, or outliers above 4000ms, including all redirects commulative.