Hi DS 
Hi cojak
This 3 modem dongly thingy is still very variable; it seems to have improved slightly recently, to the point where I can now identify times when its proxies are doing stupid things, rather than it just being generally poor service.
Sometimes it seems that its proxies between here and the Internet side have been configured with very short timeouts after which they give up, but never report this back through whatever layers lie between them and my connection, thereby leaving things hanging. This suspicion is bolstered by the fact that sometimes, one might try to reload a page that has hung, thereby initiating a totally new HTTP request, and the page will arrive almost straight away; and this from sites like the BBC.
I suspect something like:
where when 3external cn't get it from the 3cache within a certain time it just hangs, and fails to return a response of any kind. Meanwhile, the 3cache has got it, but finds that the original connection making the request has been dropped. Then, when I initiate the request again, it just comes straight from the cache and provides an (almost) immediate response.
This theory may be supported by the fact that no CUK page provides any cacheing information (no ETag, no nuffink), meaning that a full round-trip request has to be made every time; and TPD in particular can be slow at times (I must contact the admins to suggest ways to fix that - it's database-bound, basically) which, if anything between 3internal and 3external times out before 3cache can pass through the response that it won't cache, would lead to the kinds of patterns of fail that I'm beginning to identify - signal in the noise.
Give me enough failures, and I'll eventually find a consistency there; although it could be just me seeing an elephant in the clouds. Still, more data should help to make things apparent. Alternatively, let me into 3's network staff's office with a cudgel
Hi cojak
This 3 modem dongly thingy is still very variable; it seems to have improved slightly recently, to the point where I can now identify times when its proxies are doing stupid things, rather than it just being generally poor service.
Sometimes it seems that its proxies between here and the Internet side have been configured with very short timeouts after which they give up, but never report this back through whatever layers lie between them and my connection, thereby leaving things hanging. This suspicion is bolstered by the fact that sometimes, one might try to reload a page that has hung, thereby initiating a totally new HTTP request, and the page will arrive almost straight away; and this from sites like the BBC.
I suspect something like:
Code:
Me -> Dongle -> 3 -> 3internal ~~~ -> 3external -> 3cache -> t'Internet
This theory may be supported by the fact that no CUK page provides any cacheing information (no ETag, no nuffink), meaning that a full round-trip request has to be made every time; and TPD in particular can be slow at times (I must contact the admins to suggest ways to fix that - it's database-bound, basically) which, if anything between 3internal and 3external times out before 3cache can pass through the response that it won't cache, would lead to the kinds of patterns of fail that I'm beginning to identify - signal in the noise.
Give me enough failures, and I'll eventually find a consistency there; although it could be just me seeing an elephant in the clouds. Still, more data should help to make things apparent. Alternatively, let me into 3's network staff's office with a cudgel
"Recent Rain" - well, nearly right 
Comment