On Friday, August 29, 2003, at 09:58  AM, Rebecca S. Guenther wrote:

> On the second question, I was going to answer the same way that Suzanne
> did. The record describes a given resource with an identifier. The
> resource was accessed on a particular date and is being described in
> terms
> of how it appeared on that date. So is there really a need to link
> them?
> The only situation I can think of when you might want to associate the
> date with an identifier is if you are giving more than one URI and you
> accessed them on different dates-- maybe a mirror site or something. If
> people think it important to enable linking the dateAccessed with the
> identifier, we could use the ID and IDref constructs available in XML
> as
> attributes to dateAccessed and Identifier, which would provide a
> linking
> mechanism.

I don't necessarily mean literally linking them via ID and IDrefs
(though that's certainly possible).  Maybe there's another way?

I do not personally see using more than one url in a record (though I
suppose I could see have an html file and a pdf as links), so it is not
perhaps an issue for me personally.  I just think it might introduce
inconsistencies to have an access date that is by definition associated
with a specific URL, but then allow more than one URL.  In the user
guide, there is this example:

<identifier type="uri" displayLabel="Active site (if available)"></identifier>

<identifier type="uri" displayLabel="Archived site">*/</

OK, so you add:


..but there's no way to know for sure which URL(s) -- the first, the
second, or both -- the access date refers to, nor for an xslt formatted
engine to know which to use for a citation.  Yes, you could search for
the display label value, but this is not a good idea because it will be
a) specific to an institution or user, and b) prone to error.

So I guess there are two ways to look at this. One is that this is a

Another is...ah, Karen's message just popped into my box.  She said:

> I can imagine the "date accessed" being an attribute of the URL,
> rather than being treated as an attribute of the citation as a whole.
> You may also have a bibliographic date on the item.

This would be my preference.  What if the above examples were instead
something like the following?

<identifier type="uri" type="primary" accessdate="20030205"></identifier>

<identifier type="uri" type="archive">*/</

> So if I cite a web page that has a copyright date of 2002, but I
> access that page on May 3, 2003, then the date of the cited document
> is 2002, but I am authenticating the URL only as of May 3, 2003. In a
> sense, this latter date often substitutes for a date of publication
> when the document itself gives no indication of a publication date,
> but in fact it is information about the URL in my mind. The
> publication would actually be "n.d." in library parlance -- "no date."

It's not even that the access date substitutes.  If I cite a New York
Times article available online that was published in 2002, but I access
it in 2003, both dates are included, and the second is simply to say
"OK, urls are fragile, but I know for a fact this one was valid on this
date."   It often substitutes for missing publisher and place
information, which is how one often identifies from where to obtain a

BTW, this brings up another issue I raised to Rebecca privately:
shouldn't urls such as this be in the location element?  They are,
after all, references to where to find the resource.