- Understanding Link Problems
It's important to have some kind of idea of what kinds of problem we are expecting, so that we know, not only when to signal this to a user and when to ignore, but also what kinds of further tests might be worthwhile etc.
This area is one of the ones I have most doubts about. There are many different kinds of break I can visualise..
The computer at the other end just happens to be down for its yearly testing. Back tomorrow.
The data-centre at the other end suffered a meteor strike in the middle of the night. Back immediately we rebuild it..
For this we just need to wait the right amount of time for them to repair things. I expect most repairs are quick (less than a day) but there must be some kind of exponential decay.
The phone line at the other end is noisy and the connection keeps breaking.
The other system has an experimental new operating system and keeps having to be restarted.
The link to the other system is overloaded
Random broken behaviour just needs us to keep testing at different times of the day or night.. You should probably make sure that your program shedules some proportion of its tests at other times of day than when it normally runs in case some machine is up for part of the day only (down at night, for example).. but you should probably know about that anyway..
One day the debt collectors showed up and repossesed the machine then took the owner off to prison. There is no hope that any more information will ever come out of there.
The link is redirected to a new site
The target has been moved to a new site
It's possible for there to be subtle data corruption at the other end which the server there is not aware of.
A file has been corrupted
The server serves wrongly to certain other computers
the server makes random mistakes.