Easiest way i can think of explaining this is to get you to imagine a web page like a house with a home security system, where the doors and windows are the html, and the security system with all it's sensors is languages like java, .NET/C#, php, etc.
If the doors were made so that they could sound the alarm any time it thought the lock was being picked, and windows could tell when their glass was smashed on their own, yes that would be cool.
But now what do you do if you have a burglar cut a hole in the wall and enter your house without touching either the doors or windows? At that point, they have bypassed your door/window security (HTML), and all that's left protecting your house is the motion/infared sensors from the security system (php/java/c#).
This is the issue we would run into. If HTML had the added capability of preventing sql injections and nothing was done on the server side (ie: no motion/infared sensors in the house), what is left to prevent users from bypassing the HTML page and sending malicious stuff to the server directly?