Parte de un artículo de php|arch, de Eric David Benari, muy interesante por lo claro y práctico en puntos claves. A tenerlo en cuenta y a por nota.
"Shortly after most new sites are launched, the creators discover that some necessary elements have been forgotten. Pilots employ a pre-flight checklist to avoid problems while airborne, so it is only natural that web sites should have a prelaunch checklist to avoid problems once users begin visiting the site. This is more of a detailed checklist than a tutorial, but every topic in the list can easily be Googled for more info if needed.
Error Management
Does your site have a 404 Not Found page that includes links to functional pages? If not, you may lose a potential visitor who follows an old link to a page that is no longer on your site. Ideally, this should be a smart-404 page that attempts to provide suggestions as to what the user might have been looking for. At the very least, stick a Google sitesearch box on there, pre-populated with the name of the page that was not found. Applying the same treatment to the 403 Forbidden and other error pages is a good practice as well.
Valid Markup
Does every page validate as XHTML (or at least HTML)? If not, your site is not likely to display consistently between browsers and will probably be a real mess when browsing from a mobile phone.
Basic Accessibility
Does your site render consistently between IE 6, IE 7, IE 8, FireFox 3, Safari and Google Chrome? Does your site display correctly on both Windows and OSX? Is the site usable at different screen resolutions? When you increase the default browser font-size, is the site still usable? Is the site legible with CSS disabled? Is it functional with JavaScript disabled? If you have answered no to any of the above, then a percentage of your users will receive an inferior user experience.
Minimal Semantics
Does every page have a unique and relevant title, metadescription and meta-keywords? Pre-Launch Checklist by Eric David Benari can easily be avoided by following the steps in this pre-launch checklist." /> Without this, your search engine rank will suffer. Relevant titles also improve the user experience when
someone bookmarks your site or looks at the browsers’ history.
Client-Side Caching
Have you explicitly defined expires headers and disabled FileETag where appropriate? This can drastically reduce the number of requests on your web server. Apache users can do this easily via the config files, either in .htaccess or httpd.conf.
Control Search Engine Indexing
Do you have a robots.txt in your webroot? All web sites should have a valid robots.txt file even if every page on the site is meant to be indexed:
User-agent: *
Disallow:
Neglecting to have the file will bloat your web server error logs with 404 ’robots.txt’ not found messages from all the search crawlers trying to read the non-existent file.
Favicon
Does your site have a favicon file? The icon itself is of questionable value considering that it is rarely even noticed, but if you omit the favicon then each visitor to your site will add an additional 404 ’favicon.ico’ not found error message to your log files. This is similar to the previous robots.txt issue but considerably more severe; only bots and search crawlers request the robots.txt, but the browser of every visitor coming to your site requests the favicon. If you are logging to the same hard disk that serves the web site (not an ideal practice), then this will also cause a minor performance penalty as the disk needs to write to the log file for each new visitor in between serving the files on your web site.
Minify JS and CSS files
Are your external files as small as they can get? Minification is the act of stripping out whitespace and comments as well as implementing other space saving techniques such as consolidating CSS statements and refactoring JavaScript variables to minimize size (e.g. variables showAdvancedSearch and isModernBrowser might become v1 and v2). This typically reduces the file size by about 15%, although this amount varies depending on the minification scheme as well as your commenting and white-space habits. It is critical to always verify that your minified JavaScript still works in IE; the JavaScript engine in IE requires a space after certain statement types that FireFox is able to execute flawlessly with one character following the next.
Consolidate JavaScript and CSS
Are you avoiding inline JavaScript and CSS? Are you serving just one JS and one CSS file? Moving your inline JavaScript and CSS into external files allows the browser to cache the content. Consolidating them all into one JS and one CSS file also reduces the number of requests needed to load the page. This is the best-practice way of doing things, but in the real world it is not always the optimal way, especially if many of your pages have JS or CSS that is
solely used on one page of your web site. Placing JS and CSS inline can often yield a faster page load if adding a few bytes of un-cached inline code will reduce an HTTP request or considerably shrink the size of the consolidated global file.
Search Submission
Do you have a valid XML sitemap? Have you submitted your site for search engine indexing? Register your domain in Google Webmaster Tools, create the validation file as instructed by Google and then create an XML sitemap as described in the Sitemaps section of Webmaster Tools. After launch, you will need to log into Webmaster Tools and tell Google the URL of the sitemap for crawling. This will get Google to queue indexing of your new site so that it will show up in the Google search results. Next, add your sitemap to your robots.txt:
User-agent: *
Disallow:
Sitemap: /sitemap
Then submit the site to Yahoo! Directory, Yahoo! Search, MSN and DMOZ. To conquer the hundreds of other search engines, I find it easiest to use a URLsubmission service which submits the site to many engines with a few quick clicks.
Trim Down Headers
Are your server responses only consisting of valuable data? Turn off expose_php in your php.ini file, and remove any unnecessary page headers to reduce the size of each request. Also, check the site with YSlow and follow its suggestions.
Compress Throughput
Have you enabled gzip compression on your web server? Gzip greatly improves your network throughput at the cost of a negligible amount of CPU overhead.
Domain Distribution
Are the external files accessed by your site (images, JS, CSS, etc.) distributed over four domains? This may seem strange, but four is a magic-number! Many browsers, by default, will not make more than two simultaneous connections to a single domain, but it will make many connections to different domains simultaneously. If all your content is served from a single domain, then the browser backs up into a queue as it requests two files at a time. Spreading your files over many domains (and you can use sub-domains like images.yourdomain.com) resolves the issue but at the cost of a slight delay as each domain needs to be resolved before any requests can be sent to it. In most situations, four sources of data is the ideal balance between simultaneous request limitations and domain resolution latency.
RAM
Are you serving every static element of your site from your server’s RAM? The most-effective way to maximize your server’s traffic-handling capabilities is to avoid disk IO as much as possible. Serve all your static files from a RAM disk. Since RAM data will be erased if the server reboots or the OS crashes, you will need to create a shell/batch script that automatically restores the files from the hard disk upon reboot.
The following Windows DOS commands will restore files to the RAM disk R: from a HDD-based location on the C: drive.
mkdir R:\your_site
xcopy C:\your_site R:\your_site /E/H
Now, simply put the above code into a text-file with a .bat extension, and then create a shortcut to the file in your Windows startup folder. The folder location varies with Windows versions, but the path is not really needed, you can just double-click the Startup folder in the Start Menu to open the folder. MySQL users can also create tables using the memory engine which stores the data in RAM. This is ideal for data that rarely changes such as a lookup table that indexes country codes. Just like with the RAM disk, the data in a memory table will be lost upon reboot so you need to create a SQL script to repopulate your memory tables from disk-based tables.
insert into countries_memory select * from countries_innodb;
You can then set the file to execute automatically on reboot by adding the init-file directive to your my.ini file (or my.cnf on *nix). While you are editing
the MySQL configuration, you should also be sure that query cache is enabled and that the cache memory limits are set appropriately for your hardware (the default settings were created many years ago, and do not take advantage of modern computing power). For all dynamic data that is suitable for caching (both file-based data and database queries), use memcached or another caching package. With memcached, you gain the additional benefit of being able to store your PHP session data in memcached as opposed to in files (the default) or in the database (an option best avoided).
Preload Configuration Files
Are you preloading configuration files? Apache users should preload all .htaccess files and then disable on-the-fly .htaccess parsing. This saves a whole lot of file IO on your web server but also requires you to restart your web server whenever you want to implement a change made to an .htaccess file. In the section of your httpd.conf file that defines your site, remove any existing AllowOverride statements and add the following:
AllowOverride None
Include "/YOURSITE/.htaccess"
Restart the web server for the change to take effect.
Error Handling
Are PHP error messages being logged but not displayed on the page? Is error reporting set at E_ALL or higher level? Be sure that in the php.ini file of your production server, log_errors is on, display_errors is off and error_reporting is set to E_ALL or higher (ideally E_ALL | E_STRICT). This is especially important if your development and production environments are running different versions of PHP or if there are any serverrelated software configurations that are not identical between environments.
Security
Can passwords or back-end source code be accessed by simply navigating to a URL? Some common examples to check are:
• http://YOURSITE.com/.svn
• http://YOURSITE.com/.htpasswd
• http://YOURSITE.com/config.inc
SSL (If Needed)
Did you install the SSL certificate? Can you access your site via both http:// and https:// without receiving any invalid-certificate alerts or similar warnings?
If your site will need to transmit to the browser securely over the HTTPS protocol, then you will need to install an SSL certificate that is registered to your site’s domain name and issued by a real certificate authority (not self-signed). If you access your site over SSL by any means other than the domain name (e.g. IP address), then the browser will warn you about an invalid certificate. If your domain does not yet point to the server, then you will need to confirm that this warning will not occur post-launch. The way to do this is to open your operating system’s hosts file in a text-editor (on Windows this is typically found at C:\Windows\System32\drivers\etc\hosts), and add the following to it on a new line:
127.0.0.1 YOURSITE.com
The hosts file is usually loaded and cached upon browser startup so you will need to close any browser windows that are already open and then open a new browser window. Now, navigate directly to your web site via the domain name (http://YOURSITE.com), and the page should load normally. Next, verify SSL (https://YOURSITE.com), and you should not receive any browser warnings.
Email
If this is a new domain, have you set up the necessary email accounts? Verify that every email address appearing on your site is functional."
