Slashdot Mirror

Use basic HTML tags for formatting. Slashdot allows only some basic tags. See "What are post modes?" on this page for the list of allowed tags.

Crash course.

Use <p> to start new paragraph and <br> for a line break.

<quote> Quoted text </quote> is rendered like that:

Quoted text

<a href="http://whatever">Link to whatever</a> is rendered like that: Link to whatever.

You can a;ways experiment using the "Preview" button.

that something ridiculous, like probably over 75% of users don't even know what an URL bar is and I wish I was kidding. A huge amount of users don't understand the concept of URLs, domains or anything of the sort and type anything and everything (both actual search queries as well as full-blown http://whatever/ URLs) into the search engine toolbar in the top-right corner of Firefox and/or IE.

This already works since Ubuntu 7.10 (Gutsy) released over a year ago: https://wiki.edubuntu.org/AptUrl - just write a link that goes "apt:fortunes;frozen-bubble" instead of http://whatever/ and when an Ubuntu user clicks on this, it runs the equivalent of "apt-get install fortunes frozen-bubble". You can install any number of apps launched with a single click, and it's safe because they come from the repositories defined on the user's machine in /etc/apt/sources.list.

If you want to nerd out, you can bootstrap debian from floppy. Then slap down an Ubuntu sources.list and install ubuntu-desktop or ubuntu-server or what have you. If you mount the Ubuntu CD (or its image) and symlink /var/www/ubuntu to the CD root, then you can use http://whatever/ubuntu as your repository for the install. I haven't actually tried this, but my sources say yes.

Imagine I created a simple "hello world" dynamic page: something that when given a number as parameter, would return "the double of $num is $double". Imagine you would call it with http://whatever/get_double?num=10. How many req/sec would you expect on a decent machine with your favourite CGI language or web framework?

As fast as squid could return the cached pages. I would judge a framework by how easy it is to tag a controller method as being fully deterministic, and therefore generating an ETag off the request data alone, setting the expiration to never. I don't want to have to have a deep understanding of the cache architecture or do it by hand, I just want to say "deterministic" somewhere and have the framework do the rest. Bonus points for handling cache coherency for database CRUD operations or other nondeterministic operations in a way that still doesn't make me do it by hand.

Pretty much all the "agile" frameworks are 10 tons of fail when it comes to this test.

I've asked the question below in other fora, but could never find a conclusive answer; here's hoping that /. wisdom will change that...

So, when people complain about Rails being slow, just how slow are we talking about? Are we talking in the range of 10s, 100s, or 1000s of requests per second? And how does it compare to PHP, the Python-based frameworks, etc?

Imagine I created a simple "hello world" dynamic page: something that when given a number as parameter, would return "the double of $num is $double". Imagine you would call it with http://whatever/get_double?num=10. How many req/sec would you expect on a decent machine with your favourite CGI language or web framework?

(And yes, I realise just how awfully simplistic this example is, but I would like to get a ballpark figure). Thanks!

With that scheme you better make sure no one gets hold of your .bash_history file.

I use ZSH, and as far as I know there are no files that log the shell history for that. At least I hope not...

~/.bash_history is worthless in my opinion... It's pointless for auditing because a knowledgeable attacker would just run "wget http://whatever/shell.tar.gz", extract it, and have a non logging shell to work with.

Or at least

I'm surprised no one has mentioned this (or at least I didn't see it in highly-rated comments.)

Most weblinks are harmless content URLs, but prefetching certain links can actually cause a side effect - besides the obvious one (messed up weblog stats.) For instance there are plenty of web UIs that use links to delete items, log out, etc. Simply hovering over these links will cause them to be invoked.

Most links are probably well-formed, such that actions use a dynamic URL link http://whatever?action=..., and the prefetcher won't follow those. But I've seen several sites that use apparently-static links to perform actions.

Wide adoption of prefetching is going to hurt these sites much like the XP SP2 "download protection" did... Should be interesting to see how it plays out.