Log in

No account? Create an account
Linux Community's Journal
[Most Recent Entries] [Calendar View] [Friends View]

Monday, April 14th, 2008

Time Event
[OT] Dereferencing HTTP Links
Not strictly Linux related but this is a fairly active technically-minded community, perhaps someone can explain this to me.

As far as I understand HTTP client-server interaction, and the way browser user agents are supposed to behave, there is no way for a server operator to specify that he does not want referrer information sent to sites that he links to. The best you can do is use a cloaking service or the Google /url?q= trick. This will still send a referrer but it will not be the 'real' referring site.

I have seen a couple of sites that claim to have scripts on their servers to strip that information from the links they have. Unless there's some part of the HTTP specification I am unaware of that addresses this, Referer headers are a client-side behavior that is not controllable by an HTTP server. The user may elect to install software that removes this information from outgoing HTTP requests but it is not possible to derefer links from the server side, in a strict manner of speaking.

Am I wrong? Does anyone know how such a script might work, outside of using an external cloaking service?

Current Mood: curious
ExtX and reserved blocks
By default, mkfs.extX reserves 5% of a filesystem for the root user.

The purpose of root-reserved space has always been to ensure that root can login and carry out various administrative actions that might require a bit of disk space, such as touching files in /var to indicate a new login. Sometimes you have to spend a bit of space to make space.

But 5% is a huge amount, these days. On a 160GiB disk, that's 8 GiB! Even the lowest you can specify with mkfs.extX, 1%, is 1.6 GiB! That's excessive.

You can, however, turn the root-reserved blocks down after mkfs-time with tune2fs. The -r option is what you want. -r takes an argument specified in blocks, which on any FS big enough to futz with the reserved size will be 4k. So, pass to tune2fs the number of megabytes you want reserved for root * 256.

I often figure 50MB is enough for a filesystem not mounted on /. So, that's tune2fs -r 12800 /dev/path/to/device/node.

On a root FS, 500MB might be wise, especially if you have a /-is-everything partition scheme. So, tune2fs -r 128000 /dev/path/to/device/node. That's still a lot less than 1.6GiB on a 160GiB disk!

The moar you know.

<< Previous Day 2008/04/14
Next Day >>
About LiveJournal.com