So yesterday I bitched about the fact that Chrome, replaced 404 pages with a input box that would let user perform google searches. I did not like the idea , that 404 where replaced by a search box.
Today I downloaded the source of Chrome and had a look inside and found this interesting comment :
// When the frame request first 404's, chrome may replace it with the alternate
// 404 page's contents. It does this using substitute data in the document
// loader, so the original response and url of the request can be preserved.
// We need to avoid replacing the current page, if it has already been
// replaced (otherwise could loop on setting alt-404 page!)
bool is_substitute_data = loader->substituteData().isValid();
// If it's a 404 page, we wait until we get 512 bytes of data before trying
// to load the document. This allows us to put up an alternate 404 page if
// there's short text.
in http://src.chromium.org/viewvc/chrome/trunk/src/webkit/glue/webframeloaderclient_impl.cc?view=markup&pathrev=1359 around line 252.
This simply means that if you want to keep your 404 pages they need to be 513 byte in length. If they are 513 Chrome considers that the page contains interesting bits of information for the user and does not replace it with a google crafted page.