Slashdot Mirror

Yes gzip can compress. Count the number of pages that use this. Not enough. Sadly, quite few. Could more apps do this? Sure. But we don't.

Sadly, netcraft doesn't keep stats on this. They should. Mod_gzip/mod__deflate and "http compression" aren't on by default in apache and IIS, which is a shame. Also, gzip support in the squid proxy is an add-on.

Could we compress non-English fonts? You bet. Just because there's an extra 'bit' used to do things causes little compression pain. Extend that to the tough or tokenized characters sets and you still buy lots of bandwidth back.

Unicode isn't (primarily) about fonts. Browsers aren't fetching .ttf files, are they. (Actually, UTF-8 is a kind of compression-code to begin with, seeing as how non-common codepoints require more bytes than common ones - though if your page is in Chinese another encoding will be more benificial, size-wise; still I'd say go with one of the UTFs). Please do not ever again suggest using 7-bit or 8-bit encodings exclusively, in order to save on size. That's like saying 2 digits will suffice to encode the year. If you're going to go and say things like that, please invent a timemachine and go back to 1975.

When we get into MPEG codecs, you can get pretty lossy without pixelation, but again, it's not a zoned compression (zoned meaning that content like a face receives no compression while the pretty hillside in the background gets pounded by it, but your eyes focus on the human face where discernablity is needed).

MPEG already compresses out-of-focus backgrounds better than in-focus-foregrounds, because out-of-focus images leave less information in the discrete cosine transform. Also, the foreground is more likely to change, which motion detection will pick up on, so static content won't get as many bits of the stream to update its content. Still, go ahead and use H.264 if you want. The bleeding edge GPUs from ATI have it built-in.

The simple truth of the matter is that the RD bit is a useless piece of frippery, a mistake in the design of the DNS protocol, and DNS softwares should simply ignore it, whatever it is set to

I pre-install Webmin on our Linux Virtual Private Servers. I think its a great tool.

It is a great help for setting up programs that require a fair bit of configuration. After you're done, you can look at the config changes that Webmin made. Next time around you are a bit more empowered to make the changes via the command-line.

Unlike some control panel software, it doesn't 'take over' your server. It doesn't overwrite RPMS with its own custom versions. It doesn't make config files 'Webmin only'

The help from within Webmin itself is not so great. Often you have to have a very good understanding of the service you're configuring before Webmin will make much sense. There is little contextual help in the program.

I've found The Book of Webmin to be quite useful. Plus I've created a few of my own HOWTOs for common tasks:

• Setting Up DNS: BIND via Webmin
• Setting Up Email: From Sendmail to Outlook
• Host Multiple Domains: Setting Up Virtual Hosts With Webmin
• Securing Your Server: Setting up a Linux Firewall Using IPTables and Webmin

I don't know how serious it is, but check out squidng. Supposedly, it uses some kind of raw interface into reiserfs to keep the files for really fast lookups, etc.

[...]people may recall that Craig recently got in to a nasty flame war with Dan Bernstein

anyone who recognises the name Dan Bernstein will realise that this is hardly surprising. the guy is a know-it-all jerk with a severe inferiority complex, even the slightest comment about his software that could be interpreted in a less than glowingly positive fashion will result in lengthy and boring flamewar from him. that's why i've killfiled bernstein, i have no interest in reading anything that he has to say. i have no interest in running his software, either. the hassle and annoyance of integrating bernstein-ware into a working system far outweighs any of the advantages.

To make a long story short (and the flame war got ugly), Craig feels that a DNS server needs to support the legacy BIND [...]

yes, that cuts a long story very short. and conveniently throws away almost the entirety of my reasons for deciding that djbdns was worthless rubbish.

lack of backwards compatibility with bind was only one of the many reasons why djbdns sucks - lack of backwards compatibility is a common flaw with all of bernstein's crapware. as is his brain-damaged configuration style and his obsessive Not Invented Here syndrome that requires him to manically reinvent things that don't particularly need re-inventing.

Now, the only specific that Craig went in to when pointing out that he did not like my DNS server is that fact that, like Dan's TinyDNS, MaraDNS has no support for BIND's zone file format.

actually, that's not true. i didn't give ANY specific reasons as to why i think maradns is poorly written.

the reason why i think that is because of bone-headed comments you made in
http://www.swelltech.com/pipermail/mara-archive/ 20 02-November/000885.html

in particular your excuse for not doing any signal handling: "I consider signal handling a security risk. MaraDNS responds to a HUP signal by terminating; one has to restart MaraDNS to reload the configuration files."

statements like these do not inspire confidence in your ability to write secure code. before i discovered that little gem, i was recommending maradns to people who wanted an alternative to bind for non-authoritative name servers. after reading it, i can't in good conscience recommend it to anyone.

I am not saying that BIND style zone file support is unimportant. However, I think Craig should be a little more courtious in requesting this feature than badmouthing MaraDNS on the Debian ISP mailing list.

1. when did i ever request this feature?

2. if i had wanted to make more than a passing comment about maradns, i would have emailed you directly.

3. if saying "maradns isn't particularly good software, but a) it's GPL, b) it doesn't have djb's weird configuration style and c) it's adequate for the task i want to use it for." qualifies as "badmouthing" then you really ought to acquire a thicker skin.

What has happened is this: There are some people who got in to an argument with me in one of the journals; some of them didn't like the fact that I won the argument against them, so are now trolling me.

Note that they do not mention which test queries return incorrect IP addresses; and also notice that they way they make it as personal as possible.

Anyway, I have this testimony that MaraDNS is an excellent piece of code.

Note also that this troll does not have the courage to tell us who they are, since I would promptly put them on my foes list.

I probably should not feed the trolls by replying to this, but feel that I should clear things up for people who are reading this.

Most likely, the above troll was posted by com2kid; he is mentally ill and really needs psychological help.

- Sam (posting anonymously since something like this only needs to be seen by people posting the above link)

Then again, I've had my book available for free online since I wrote the first 30 pages or so. And I intend to keep the free version available after it is printed by No Starch in a couple of months. I guess I'm one of them weirdos who thinks content creators shouldn't extract money from their customers at every possible opportunity. I like musicians who sell their records for reasonable prices (Dischord records, That Dog before they split, etc.) and authors who don't really think everyone should have to pay $45 for their books (Bruce Eckel). I'm all for getting paid, and paying folks who have done the work, but I'm also for letting people decide if a work is worth shelling out the dough for.

I guess this Open Source thing has gotten to me. Or maybe it was the punk rock thing ("give the kids their moneys worth"). Regardless, prices on new books are probably higher than they ought to be, and if I often buy used ones if I can't get it new for a reasonable price--the new price always dictates whether I get it at my local Half Price Books or at BookPool.

That's my take on the issue anyway.

They do not try to be the lowest price in the market, and many even charge AOL style prices for their dial up service. But, they do everything they can to insure that their service 'feels' faster and better and friendlier. They always have a real person answer the phone. A lot of times it's a two or three man show--but they answer their phones when you call with a problem and they look into the problem right then and there--no trouble tickets, no "I'll email the tech department and let them know". They pursue business customers (but still treat their consumer customers as old friends), because of the much higher margins, and they know many of their clients by name (or they keep a database so at least they know of all of the old problems and complaints and issues the client has had).

As has been mentioned, small towns are still ripe for entry by ISPs. Even in the US. Several of our customers are doing great, and they are still the only ISP in their given market. The ones who are seeing competition are staying ahead of the curve by offering business and consumer wireless web service (the last great frontier of the ISP world--read some wireless ISP mailing lists for a really fun bunch of folks--erecting towers without licenses, installing broadband in the most bizarre places you can imagine, and overall working together in a beautiful coopetition banding together against the Bell's). Wireless broadband is very cool in that if you really want to, you can push as much as 11Mbits (usually only 2Mbits, however) across town.

And of course, I also think a good web cache is a great idea for a small ISP. Bigger ISPs may have bandwidth and money to burn, but the little guy has to save where he can--and provide better service to boot. The original post asked for prices...starts at $1500 for a build it yourself box running Squid--here's the directions:

http://www.swelltech.com/pengies/joe/squidtuneup/t 1.html

Or you can buy one from us or other, proprietary vendors. Our Squid based appliances start at just under $2k to support up to two T1 lines, and there are quite a few other caching vendors who are quite happy to work with small ISPs. Stratacache is a Volera-based vendor that has some low-end options. The company seems to be run by good people, so is also a good choice if you don't mind a proprietary solution.

You should further look a little deeper into the results. Microbits box was only caching about 44% of web traffic and getting rather slow response times. So while they got 120 reqs/sec, no sysadmin in their right mind would push that box that hard. To compare apples to apples with the Squid entry or the Swell entry (both had nearly ideal cacheability and excellent response times) you should think of the Microbits box as being more along the lines of 95 or 100 reqs/sec.

To see Squid results in more favorable light, check out the more recent results on the Swell web page:

http://www.swelltech.com

Our test box at the bake off was having fits using async io...so we disabled it in order to get a clean run. However, performance suffers markedly without it. Those async issues have been resolved...Our boxes are running in our labs at 110 reqs/sec right now (we have a 100 reqs/sec run benchmark online...you can note that response of squid is still excellent at that load).

Anyway, given the proper tweaks, Squid can really scream on a low priced box. (Our $2139 unit is the one included in the bakeoff and our more recent benchmarks.)