In short, I was expecting this to be far easier than it turned out. I don't think I would be wrong in saying that this was a more-challenging paper than any of the 2003 or 2002 NSA exams. One particular sore point was that it was fifteen marks longer than its predecessors from 2003 - so the two couldn't be compared like-for-like, and practising an old paper to get the timings right was worse than useless. I'm not entirely antipathetic to the extended paper - 60 marks in an hour and three quarters is probably underestimating our abilities somewhat - but it would have been nice to have some forewarning. Anyway....
Section A was on the third lab exercise, as many correctly predicted. The standard sixteen marks were for a description of the requirements, a diagram of the design, and details of the sockets calls and parameters. The old testing chestnut was next for four marks... room for some creativity there. Sadly missing was the question that compares one's solution to the sample, which is usually good for a few marks. In addition to the staples, there was a question which asked us to consider rewriting the client using UDP instead. Hmm, well not having done any UDP sockets programming in C, this wasn't quite easy, but I managed to string a few points together. I'm not entirely sure whether some other calls would have been absent, or whatever, but I did draw a pretty diagram, if that's any consolation.
Question B1 was about protocols, services, connection-orientation, and subnets. Mostly straightforward definitions, a 12-marker on {C-O, CL} X {Protocol, Service, Subnet}, which wasn't entirely pleasant, and the repetition involved may have induced dizziness in some. Also two R/A questions on layering protocols and services, which were tricky.
Question B2, was on LANs, with some freebie definitions, a discussion of CSMA (for 10 marks!), and a tricky question which forced us into giving a definition of reservation, which never was adequately (3 marks-ly) defined during the course. And one little bit of reasoning about how collision detection would affect the choice of CSMA persistence. Fairly fair, I'd say.
Question B3, which I eschewed, was about flow control, congestion control, error control, and the contrasting roles of the Data-Link and Transport layers. No massive questions there, if the sight of double figures makes you squeamish, but it wasn't my best subject, so I stayed away.
So, in all, a paper not designed to leave you with a warm fuzzy feeling. Let's hope we're saying the opposite tomorrow!
Cheers,
Derek.
| 
home
blog
dreams
photos
ducks
pastry
