rgrig / homework-eval Goto Github PK

The dialog that appears after submitting must say clearly what was the result 
of the evaluation.

It should distinguish between WA and LE, in a way that fits with multiple tests.

Setting up should be automated in a script.
That script should be put in a debian package.
There should be a target in build.xml that produces the .deb file.

1. Make the session last for more. (I think this means use a cookie instead of 
HTTP sessions.)

2. If the session expires and a solution is submitted, then give a good error 
message, not the current "does not compile."

The code with collections is getting really ugly.

1. Focus should be initially on the 'psudonym' box.
2. Enter in the password box (and in the pseudonym box?) should behave like 
clicking Login. Perhaps I should even ditch the login button.

Use GWT's recommendations.
Make sure it looks nice.

Server crashes on infinite loop with a printf

Distinguish the "Score" tab somehow.

Since this software is supposed to be used for various purposes, it should be 
easy to configure he title of the site. And perhaps allow some other kind of 
branding.

The colors are a bit messed up and certain things are hard to read. For 
example, the title of the dialog that appears after submitting a problem.

The math looks nicer when formatted by MathJAX.

Either remove everything related to quizzes, or check their code and clean it 
up.

I seldom use quizzes, so I think it's not worth maintaining.

This uses time penalty as a tie breaker.

The penalty for a problem is the time it took to solve (in minutes) plus a 
constant penalty for each wrong submission.

I believe the "submission time" recorded now is really the time when judging 
starts, after perhaps staying a bit in a queue. Since penalties are used for 
ranking, this should be fixed.

Tasks are presented in some random order.
They should be in some definite order.
The easy ones should come first (add "difficulty" as a task property).
Then use some tie-breaker, perhaps deadline followed by task id.

Use Window.alert ONLY to diagnose bugs in the judge.
Otherwise avoid, because it freezes the whole browser, which is annoying.

It should be possible to use an external program as a validator.  That would 
allow, for example, tasks with more than one correct answer, or even 
optimization tasks.

Make it possible to set a time where the scoreboard is frozen.

Introduce different scoring methods, proportional and binary.  The current 
behaviour would be reproduced by adding
  scoring proportional
to the main config file.

Changes to the configuration are seen by clients only after logging-in again. 
They should either:
(1) poll the server for changes, or
(2) get a ping from the server when something changed.

At the moment the order of output on sample input is not fixed. Needs to be 
alphabetically sorted.

It's getting silly to keep converting to arrays.

Make sure there's no deprecation warning.