Can one of you lovely gentlemen explain the difference please?
I know what *I* meant, but I'm not sure what you all mean.
*smiles*
<snip>
In English the meanings of words, labels and terms are determined by no
more than what people use them to mean. And very few words or terms have
a single distinct meaning anyway. Meanings change with time and usage
and, while you can regret the corruption of the meanings of word as they
change, there is nothing that can be done to stop it. People also
suggest and invent new terms, some of which may be seen as preferable to
those currently in use and so become more widely adopted
This doesn't really matter, as the important thing is being able to
communicate. So, so long as I understand what is meant communication has
successfully happened, and if not I can ask for clarification.
The insistence upon the application of over-literal rigid definitions to
ordinarily communication (as opposed to technical specifications) that
characterises, for example, many of Thomas Lahn's posts, seriously gets
in the way of communication. Partly because arguing about the meaning of
words in English is infinitely recursive (as all words are defined with
words, which can themselves be subject to the same arguments), but
mostly because people expect some level of co-operation in understanding
from the person they are communication with and get frustrated when the
recipient isn't willing to try to understand.
This makes me unwilling to insist that any particular set of terms
represent the true and correct labels for use in this context. When I
read regulars on this group suggesting "object detection" I know what is
meant, because I know they are not advocating what I would currently
term 'object inference'.
However, I do have a strong attitude toward which terms are best suited
to the various activities and strategies that apply to browser
scripting. So I apply that attitude to what I write on the subject and
would leave it to others to make up their own minds as to whether my
preferred terminology best suits the situation.
The primary context for the application of this terminology is the
handling of the variations in the execution environment of browser
scripts. That is; the strategy used to cope with the fact that an
Internet browser script must be written without any foreknowledge of the
environment in which it may be executed.
The attempted strategies have been:-
1. Ignoring to problem and writing for some individual
browser/browser
version(s) + browser configuration(s) that appear to satisfy some
ill-defined (usually "statistical") criteria.
2. Browser detecting; by UA string or by object inference
(sometimes
termed 'object detection' in this context).
3. Attempting to match the behaviour of scripts to the specific
capabilities
of any web browser by examining the execution environment to see
if
it provides the facilities that the script would like to employ.
The third strategy is currently the preferred approach to the creation
of cross-browser Internet scripts. The first obviously never results in
a cross-browser script and the second is demonstrably unreliable to the
point of being useless, while the third (properly applied) can get the
maximum out of any browser environment while doing the most to
facilitate controlled 'clean degradation' (another term we could argue
about the exact meaning of).
Applying labels to those strategies is convenient and useful when
talking about them. We don't really need a label for the first, though
it might be refereed to as "Intranet scripting" (the alternatives being
somewhat unkind ;-). The second can be lumped together behind the term
"browser detection" (as that is a well established, self-explanatory and
fairly obvious term to use). The third being variously labelled "feature
detection" and "object detection".
My preference is to apply the label "feature detection" when talking
about the *strategy* for handling the differences between browsers. I
prefer it in part because the actual testing done in order to implement
the strategy is not restricted to objects (or even objects + methods,
which are both objects in ECMAScript terms anyway), but mostly because
searching the internet for "object detection" will turn up web pages
that describe the problems with browser detection by user agent string
and then go on to advocate browser detection by 'object inference',
applying the label "object detection" to that approach. This makes me
think that a distinct term is needed for a distinct strategy.
The next level of the application of the terminology under discussion is
to the testing itself. And this is where the term 'object inference'
enters the picture. As I recall it was Lasse Reichstein Nielsen who
first described an example test as "object inference", and I like that
term as it implies that the test is indirect. However, 'object
inference' features in both the 'browser detection' and 'feature
detection' strategies. It is the wrong approach for 'feature detection'
(as a strategy for handling browser differences) as there is always a
more direct test available.
So we have:-
isIE5Pluss = Boolean(document.all && document.getElementById);
- which is 'browser detection' by object inference, and:-
if(document.images){
img = new Image();
img.src = 'some.gif';
}
- which is an attempt to match the execution of code to the environment,
so a 'feature detection' strategy, but it is a test that infers the
existence of one feature from the existence of another.
Unfortunately a search of the Internet for "object detecting" will also
turn up pages that talk about the unreliability of browser detecting and
propose 'object detecting' as an alternative, but then go on to list
examples of this type of object inference testing as object detection.
The 'correct' alternative test would be something like:-
if(window.Image){ // or:- if(typeof Image == 'function'){
img = new Image();
img.src = 'some.gif'
}
- and I would label that 'feature detection'. It is a test of an object
(assuming functions are to be classed as objects, as per ECMAScript's
internal attitude towards them) but the distinction that I see as
significant is that this test has a direct one-to-one relationship
between the test made (the object tested) and the code that is to employ
the result of that test.
"Feature detection" is also a good term to be used to describe the test
used to implement a "feature detection" strategy in cross-browser
scripting because testing can involve subjects considerably more diverse
than just objects (including or excluding functions and methods).
Code such as:-
if(typeof normalisedStyleObject.display == 'string'){
// assume the switching of the display property is
// available in this environment.
}else{
// don't expect this environment to support display switching.
}
-or:-
if(!('a'.replace(/a/, (function(){return '';})))){
... //function references OK with String.replace method.
}else{
... //no function references with replace method.
}
- are tests of a browser environment to examine its level of support for
something, that could not be easily categorised as 'object detection'
tests as they do not test objects at all. Rather a property in the first
case and a method's behaviour in the second. They do qualify as tests
where the environment's support for a particular feature is examined in
a way that is directly related to that feature. Making them well suited
to an implementation of a 'feature detection' strategy, and so 'feature
detection' tests.
Taking 'feature detection' as a label for tests suited to the
implementation of a 'feature detection' strategy, 'object detection'
(and 'method detection', and others) could be regarded as appropriate
labels for sub-sets of 'feature detection' tests. Though I would still
be inclined not to use the term 'object detection' in this context
because of its use in the other contexts previously mentioned.
When I use these terms I mean something along the lines of the
following:-
Browser Detection.
A strategy for handling the differences in
the execution environments of web browsers where an attempt is made to
match the script code executed to a set of known environments by
identifying the browser in which the script is executing.
Originally attempted through the examination of properties of the
browser's 'navigator' object, particularly the 'userAgent' string but
because those properties ceased to discriminating as more browsers came
into existence the thrust of the browser detection strategy shifted to
examining the browser's environment for features (objects) considered to
be indicative of a single browser type (see: Object Inference).
Currently deprecated as a practice because it is demonstrable that
scripts do not have access to sufficiently discriminating information to
allow accurate browser identification, and/or scripts authors do not
have knowledge of the range of browsers necessary to accurately identify
features that could be genuinely indicative of particular
types/versions.
Feature Detection.
1. A strategy for handling the differences
in the execution environments of web browsers where an attempt is made
to match the script code executed to the browser's ability to facilitate
its requirements by the application of direct testing to various
features of the environment needed by the individual script.
Currently considered the only viable strategy for the handling of
the differences between browser environments that is required by
cross-browser scripting.
2. Any test applied to a property, method,
object, etc, the behaviour of such, or the outcome of the use of such,
that is used to control the execution (and clean degradation) of
client-side javascript code, where the test performed has (to the
maximum extent possible) a direct (preferably one-to-one) relationship
with the code that would be executed as a result of the test.
Object Inference.
A style of testing in which the behaviour
facilitated by a client-side execution environment (or aspects of that
environment) is inferred from the results of tests made on a limited
sample of features of that environment, or on features that are only
related to those subsequently employed in the script by an observed
coincidence in a limited set of browsers.
Often employed in Browser Detection or as an (Inappropriate)
implementation of Feature Detection (1) which fails to satisfy the
desire for Feature Detection (2) tests to have a close/direct
relationship with the code executed as a result of those test.
Given those definitions a definition of 'object detection', based on the
various ways in which the term has been used might go:-
Object Detection.
1. A strategy for handling the differences in
the execution environments of web browsers where an attempt is made to
match the script code executed to the browser's ability to facilitate
its requirements by the application of testing to various features of
the environment (see: Feature Detection (1)).
2. A style of Browser Detection in which the
type and/or version of the browser is inferred by testing for a limited
set of features considered indicative of that browser (see: Object
Inference and Browser Detection).
3. Any test applied to a property, method,
object, etc, the behaviour of such, or the outcome of the use of such,
that is used to control the execution (and clean degradation) of
client-side javascript code (See: Feature Detection (2)).
4. A sub-set of Feature Detection (2) testing
in which the subject of the tests are objects (possibly including
functions/methods as those are Objects in ECMAScript).
5. A style of testing in which an aspects of
a client-side execution environment is inferred from the results of test
made on a feature of that environment that is only related to any
subsequently executed code by an observed coincidence in a limited set
of browsers (see: Object Inference).
Given such a wide ranging possible definition I can know what is meant
when I see the term 'object detection' used (by individuals who I know
to understand the issues sufficiently to be using definitions 1 and 3)
but it concerns me that individuals without previous experience who
attempt to research the subject on the Internet might encounter
information that would turn good advice into misdirection. So I won't
use 'object detection' to describe anything, and instead I will employ
terminology that is more descriptive (and/or less abused).
However, the observation that 'object inference' is a relatively
recently coined term in this context, and an appreciation that the
labels applied are ultimately nominal, leaves me open to suggestions of
other, and better, terminology to apply to our context.
Richard.