eiffelroom - information - Comments

It could be still shorter in some circumstances

colin-adams — Wed, 21 May 2008 12:03:44 -0700

That looks fine.

In fact, you can drop the eiffel: too in some cases. The application can set a base URI of eiffel: .

Colin Adams

No plan for now.

ted_eiffel — Tue, 20 May 2008 20:03:23 -0700

Thanks for the information, Colin.
I read more about this. Now I think "//" may not be needed. Since it is used ONLY when the syntax of the URI's <scheme-specific-part> contains a hierarchical structure. system/target/group/cluster/class/feature is kind of hierarchical structure, but not strict. Some parts are optional when certain parts present. It seems the best fit of current requirement is like "eiffel:?class=XXX&feature=XXX".

Are you going to register the Eiffel URI scheme?

colin-adams — Tue, 20 May 2008 00:17:25 -0700

I am puzzled as to how you are implementing these incoming links.

First, you talk about entering them in a browser. And the syntax looks like a URI but isn't.

I would suggest the following actions:

1) Read RFC 3986 - the definition of URI syntax. 2) If you still can't see why the examples are wrong, and why the design of the URI is poor, create a UT_URI object with one of those example "URI"s, and then stop in the debugger at the end of the creation procedure, and examine the attributes of the UT_URI object. 3) Read the procedures on registering a URI scheme (scheme - not protocol. Even the http URI scheme doesn't always mean the http protocol will be used). Even if you don't submit the scheme for registration (and I certainly think you should do so), considering the process should help a lot.

Colin Adams

The following link is

ted_eiffel — Sat, 17 May 2008 21:11:56 -0700

The following link is sufficient to locate {STRING_8}.is_equal
eiffel://system=ec&target=base&class=STRING_8&feature=is_equal

By your means, it would be:
eiffel:///bysystem/ec/base/STRING_8/is_equal

Note that `cluster' is omitted here. It means that we don't know what `STRING_8' is. It can be a cluster name and `is_equal' is a class name. So this could be interpreted as class `IS_EQUAL' is required in cluster `string_8'.

Path URI

nenieorg — Fri, 16 May 2008 08:01:34 -0700

Yes path URI sounds so much better than the name value mess, and is not exclusive with defaulting of various types, with say a magic word as first component:

eiffel:///byclass/STRING/is_equal eiffel:///bycluster/foo/STRING/is_equal

etc

Though I am not aware of

ted_eiffel — Wed, 14 May 2008 07:28:24 -0700

Though I am not aware of what kind of software processes the URI and how it would be useful picking up name/value pairs, it is would always be better to follow standard. Using eiffel:///? is OK for me, except that it looks uglier. Actually the .ecf is not needed for EIS incoming mechanism. Only optional system/system UUID is needed for locating project.

Perhaps a path

colin-adams — Tue, 13 May 2008 07:35:48 -0700

You would need eiffel:///?... if you expected generic URI processing software to pick up the name/value pairs.

or alternatively abandon the query parameters approach, and have a structured path (something like /system/target/cluster/class/feature). But that would make it hard to default the system and target.

I can't see how this is supposed to work anyway. Surely you need an ecf file? In which case the obvious syntax is:

eiffel:///path/to/my.ecf?class=my_class&feature=my_feature

and the authority field is then available for a future extension for non-local access.

Colin Adams

Indeed. I think eiffel:///

ted_eiffel — Tue, 13 May 2008 05:04:22 -0700

Indeed. I think eiffel:/// would be nicer.

Invalid URIs

colin-adams — Tue, 13 May 2008 03:18:33 -0700

The syntax of the eiffel: URIs is invalid.

That is, the presence of the // indicates that what follows is an authority segment:

According to RFC 3986, this is terminated by the first /, ? or # character.

None of those is present, so the remainder of the URI needs to match the authority production syntax. As there is no @ or : present, this in turn means matching the syntax for host. Clearly it does not.

I suggest that you appear to be parameter/value pairs, and that therefore this remainder should actually be part of the query segment.

Accordingly, change eiffel:// into eiffel://? in order to conform to URI syntax.

I assume that by protocol link you actually meant URI scheme.

Colin Adams