| iMatix home page
| Xitami home page
| << | < | > | >>
Xitami Xitami
Version 2.5b4

 

A Beginner's Guide

Setting-up a Web Site

General Remarks

Xitami does the same work as any other web server (only faster and more reliably), so this section covers general information that you can probably find on the Net, in a hundred books, and in documentation for other web servers. Since you chose to look here, we'll feel free to mix our opinions with our advice.

You must of course have a good idea of 'Why?' before you start building a web site. Who is going to access it, how often, and to get what information? A web site is basically an exercise in publishing. So be prepared to spend a lot of time writing and editing material. Web sites that look like video games may be fun to build, but are usually painfully slow to work with, and don't necessarily add any value to the information you're presenting.

There are many tools that help with the process of building and managing the many HTML files you will need. However, there is no substitute for a good knowledge of HTML (which is a simple language) and for some skill in managing complexity. The shareware HTMLib library is an excellent reference for the HTML language. (Htmlib is written by Stephen le Hunte: to find it, search altavista or any of the other big search engines.)

You may find that a tool like MS FrontPage is ideal for managing this problem. You may alternatively prefer a more mechanical solution, such as the htmlpp preprocessor that we use for our web site. Of course, we recommend htmlpp. It's simply more open and flexible than any do-it-all environment like FrontPage. Whatever choice you make, these are some of the issues you will have to manage when you start producing dozens, then hundreds of HTML files:

Top Ten Things To Do

  1. Learn HTML - use a reference like HTMLib to keep up to date. (Htmlib is written by Stephen le Hunte: to find it, search altavista or any of the other big search engines.)
  2. Write HTML that works with all browsers, including text-only browsers.
  3. Learn the basic rule of publishing: keep it clean.
  4. Keep your web pages simple so that they load quickly.
  5. Your home page should fit on one screen.
  6. Use an HTML validator tool to check your pages.
  7. Use a good tool to manage the web site files.
  8. Keep a test site, and test well before you publish.
  9. Use the Xitami alias functions to access other resources such as HTML-driven CD-ROMs.
  10. Use the Xitami errors.log file to detect and fix link errors.

Things To Avoid Like The Plague

  1. Blinking text.
  2. The bells and whistles offered by proprietary extensions - these are designed to lock you and your clients into vendor-specific solutions.
  3. Lots of images, unless you are building an intranet site. Images can be very useful, but cost a lot in terms of preparation, maintenance, disk space, and network transport.
  4. Java, JavaScript, ActiveX, PerlScript... unless you are very aware of the costs and benefits involved. All programming is expensive, and executable content is particularly costly. not least because it relies on untested, rapidly changing technologies. Like images, executable content is most often used for flash, not to solve specific problems. JavaScript, in small quantities, is probably the safest route to go for small adornments, since most browsers can handle it these days.
  5. CGI, unless you really need it and are aware of the costs and benefits involved. Badly-written CGIs will slow-down your entire site. Good CGIs can provide a very useful level of interactivity, but you must know what you are doing.
  6. Cookies, unless you really know what you're doing. Sites that use cookies can be viewed negatively by users. If you use cookies, make sure they are set to expire correctly, and do not send a new cookie with each page.
  7. All web servers except Xitami - why make your web site run slower? :-)

Installing Xitami

Xitami is quite simple to set-up -- basically it runs with no configuration at all -- but your TCP/IP set-up must work first. You can start by making a stand-alone site (a browser talking to Xitami on the same machine), then connect your system to a network and let other people access your pages. This is a checklist of things to do:

Getting Yourself Connected

This section describes how to make your web site available to other people. The problems involved in connecting on an internal network are a little different from connecting to the Internet itself.

How Do Domain Names Work?

So long as you have TCP/IP installed and running you can use 127.0.0.1 and 'localhost' (which both mean the same normally) and which TCP/IP interprets as the 'loopback address', i.e. the current machine.

When you want to talk to other computers you need to know their IP address. Correspondingly your system also needs an IP address. And your address has to be unique, network-wide. When the network is the Internet, this means world-wide. There are two ways to get such a unique address. One, you ask/pay someone for a fixed address. Two, you work with an ISP that owns a pool of addresses. This is typically how dial-up PPP connections work: the ISP will lend you an address for the duration of the connection.

Now, if you dial-up on a PPP line, and you do 'ping xxxx' where 'xxxx' is the name of your system, ping will tell you your current IP address. Other people can, then, connect to that address and access your web pages. This works. But they have to type the literal address: 'xxx.xxx.xxx.xxx'.

TCP/IP uses a system called DNS to translate a name like 'imatix.com' into an address. DNS uses a network of servers that are able to translate names into addresses. So when you use Netscape to access imatix.com, your local TCP/IP interface asks its local DNS server to translate the name. This goes off, and after several hops comes back with the address, and then you can connect.

The system name must of course be unique, within the network, or world-wide on the Internet. Again, you can get unique names in several ways. You can extend an existing domain name (research.imatix.com) or invent a new domain name (some-thing.com). Domain names must be registered with the Internic. This costs $70 for 2 years and $35 per year after that. You can do this directly or via your ISP. It's quite a fast process; the only problem is finding a good domain name that's not already used.

In general, you'll find that a dial-up PPP connection is not much use for a web server, since your IP address changes each time you dial-in to your ISP. (Ignoring the fact that local phone calls still cost quite a lot in many parts of the world.) There are some interesting sites that help get around this problem by acting as DNS/proxy servers for such connections. If you want to do this kind of thing, you'll have to investigate yourself -- we're getting too deep for a beginner's guide.

On An Intranet (LAN)

See your network administrator to get an IP address. Then, configure your TCP/IP software so that your computer is reachable from others on the network. Generally you'll already have got TCP/IP working even to run Xitami on a stand-alone system. What changes with respect to an intranet is that you need an IP address, and it can't be the same as any other IP address on the network.

When TCP/IP is working correctly (even before you start Xitami) you can use the 'ping' command from other machines to check that your machine is addressable. Type ping like this:

ping xxx.xxx.xxx.xxx

Where 'xxx.xxx.xxx.xxx' is the IP address of your machine. If this works, then you can start Xitami and access it from a browser using the URL http://xxx.xxx.xxx.xxx/.

The next step is to get your system known by the DNS (domain name system). An intranet usually has one or two domain name servers, although it's not mandatory. If DNS is working, you can configure your TCP/IP to talk to the DNS server - then other PCs will be able to refer to your system by name, not IP address.

If you do not have a DNS server, you can generally use the 'hosts' file (in Windows this sites in C:\Windows) to translate IP addresses into names and vice versa. Each computer that want to access your web site needs to put a line in its hosts file. This is a bit tedious, which is why DNS exists.

On The Big Wide World Wide Web

Getting your site onto the web involves cooperation from a commercial provider of some sorts. You can get help from:

A good solution for a small web site is to pay for a fixed line. There are various technologies; ISDN, ASDL, POTS (plain old telephone something) ... Check-out the capacity of the line and shop-around for the best deal. This gives you the most flexibility and control of your system, but may be limited when handling large volumes. The most interesting type of connection may be the kind offered by TV cable companies: in some cities this is a very cheap way to get high-speed IP connections. However, check the rate at which you can send data out from your server - this is sometimes quite low.

The actual steps involved in setting-up an Internet host are (and feel free to correct me on this; this is coming from long-unused memory cells):

  1. Invent a domain name that no-one already uses. You can telnet to internic.net (I think) and query their database to see whether your name of choice has already been used. You can also use the 'whois' command to see whether the name is used.
  2. You have to submit a domain name registration request to the Internic.
  3. At the same time you have to be able to refer to two DNS servers who can handle the domain name translation.
  4. Once the name is registered and the new IP address is given to the two DNS servers, it gets distributed through the Internet, and after a few days anyone can access your system either using its IP address or the domain name.
  5. It's only really worthwhile doing if you need to aquire a domain name for business reasons, or you have a permanent connection to the Net.

This is a lot of fuss for normal people, and an ISP can usually do the whole job for you, though they will charge something extra.

On a private dial-up network

You can also become your own ISP by setting-up a pool of modems, and arranging for dial-up accounts into an internal intranet. This can be very effective for networks with a specific set of clients - for instance, salesmen who travel a lot. It can also work in regions where real Internet connections are expensive or not available. If you want to do this you should find someone who knows about such things.

Using Virtual Hosts

Virtual Hosts are a useful way to manage independent and separate web sites on a single system, with one copy of Xitami running on the main HTTP port 80. The user sees separate web sites - you need only manage one server. This is of most use when you want to host several sites but only manage a single web server. You can also run one copy of Xitami per site. Since Xitami is small and does not use much memory both approaches are practical.

The virtual hosts section gives a detailed explanation of what virtual hosts are and how to set them up. We'll cover some more introductory topics here.

When you define virtual hosts, each virtual host can have its own webpages directory, CGI directory, log files, error messages, password file, timeouts, etc. In fact, almost all Xitami options except those that affect the whole server (such as the port) can be specific to the virtual host.

To define a virtual host you define a specific config file - this contains all the options that are specific to that virtual host; other options are then inherited from the xitami.cfg and defaults.cfg files.

To create virtual hosts, you must be able to define new entries in the domain name system (DNS) or be able to define multiple IP addresses on your system. Neither of these are jobs for beginners, so if you've not done it before, get competent advice.

Managing Your Web Site

Updating The Site

It's a good idea to work with a 'test site'. This is simply a directory on your PC where you install and test the HTML files, images, CGI scripts, and other resources before you install them on your public web site. For instance, the iMatix site is built on its own disk partition, where each directory matches that on our web site.

We do not generally work directly on the files in the test site. Rather, we build a package of HTML files, images, whatever, then install them into the test site. This lets different people manage different sections of the site. It's also a natural way to work when one uses tools like htmlpp.

There are basically two ways to update a web site: the 'dribble' and the 'stomp'. Dribbling means updating it in small pieces; a few files here and there. This is typically how people work when they don't use a test site. Stomping means shoving several tons of stuff onto the site at once, so that everything is updated together. Dribble works for spot updates, bug fixes, and such. But it is not a good way to work in the long term: stomping is safer and not much slower.

We stomp our site using a couple of Perl scripts that find all files changed since the last stomp; these files are compressed into a zip file, which is sent by file transfer to the web site. There we decompress it. It's a lot faster to do this than to transfer the individual files one by one (firstly, zip files are compressed by about 75% unless you are already handling compressed data, and secondly, it takes a second or two to negotiate a file transfer, which is slow when you transfer dozens of small files).

There are of course many free and shareware tools (such as Netload) available to do this kind of thing, but none that we know of will use a zip-style compression to save upload time.

Counting Hits

Xitami produces standard NCSA-style log files that can be read and analysed by most log file analysers. People often misuse the term 'hits' to imply that one hit is one person visting the site. This is not true. For instance, the iMatix web site has 250,000 hits in a typical month, but about 15,000 actual visitors, or whom perhaps 2,500 stay long enough on the main page to trigger the page counter. Each page has several images as well as the HTML text, and people will read several pages.

To accurately count the number of visitors to your site, you can count the number of hits to the main page. If you encourage people to always visit your site's main page (publish just that URL), then your statistics will be more accurate.

Using The Log Files

This section still needs to be completed.

Using Password Protection

This section still needs to be completed.


| << | < | > | >>
| Welcome To Xitami | Table Of Contents | Installing Xitami | Administration | Configuration | Using The Common Gateway Interface (CGI) | Server-Side Includes (SSI) | Using Filters | Server-Side XML Processing | Server-Side GSL Scripting | Image Maps | Virtual Hosts | The FTP Service | The Dynamic DNS Feature | Throttle Pipes | A Beginner's Guide | Writing Web Server Extension (WSX) Agents | Extending Xitami with External Peer Processes | FAQ | Getting Support | Credits | Release History | License Agreement
iMatix
Copyright © 1996-99 iMatix Corporation