"e n | c k m a" <bo*@marley.com > wrote in message
news:7F******** *********@news-server.bigpond. net.au...
<snip>
Also: "The detection of these older browsers is done by object
detection,"
I think it's referred to as "method detection", isn't it? You're
not really detecting objects, but methods.
The most common alternative (and more precise) term for "object
detection" is "feature detection" as it takes into account that some of
the things that need detecting are not necessarily objects but may get
down to ECMA Script implementation details. Such as checking to see of
the String.replace method will accept a function reference as its second
argument (as is very usefully the case in later versions) with code
like:-
if('a'.replace(/a/, function(){retu rn '';}).length == 0){
//function references OK.
}else{
//no function references with replace.
}
(it works because if only string arguments are accepted as the second
argument the function is type-converted into a string and that string
has a length longer than zero.)
However, the code in question does not use object/feature detecting in
the sense that the term is used in connection with cross-browser
scripting. It uses a technique that is best categorised as "object
inference". Testing for the existence of one method, in this case the
document.getEle mentById method, and then assuming that the presence of
that feature can be used to infer the existence of other features,
dynamic support of either innerHTML or the W3C DOM interfaces to the
level of text nodes.
The assumption that the existence of the document.getEle mentById method
can be used to infer that the browser has full dynamic display
capabilities falls down in the face of, for example, Opera 5 & 6 and Web
Browser 2 (on the Palm OS (based on NetFront)) which both implement the
method but do not implement either of the required interfaces. (Opera 6
puts up a JavaScript error message and Web Browser 2 just fails
quietly).
A feature/object detecting script would have gone to the effort of
verifying the existence of the required dynamic interfaces before
attempting to use them. Not that knowing that an element has a text node
as its firstChild necessarily means that writing to the nodeValue
property of its firstChild will result in changes in the display, but at
least knowing that the firstChild property is a reference to an object
prevents the attempt to write to one of its properties from putting up a
JavaScript error report.
The general principals of feature/object detecting is to verify the
existence (and possibly behaviour) of any feature that can be verified
prior to using it and make as few assumptions as possible. In that way a
feature/object detecting script can exploit a browser up to the browsers
ability to support the script and still be in a position to cleanly
degrade under its own control in the face of the absence of the required
features without ever needing to be interested in any of the specifics
of which browser it is that it is executing on.
On the other hand, object inference, to be effective, requires a
detailed knowledge of the DOM/ECMA Script implementations of every
version of every JavaScript capable web browser, knowledge that nobody
sane could reasonably claim to posses.
Richard.
