On Sat, Sep 11, 2004 at 05:08:04PM +0200, Marcello Perathoner wrote:
Greg Newby wrote:
Project Gutenberg does not encourage deep linking to our Web site and, in some cases, has actively discouraged it. But a link to the main page, http://www.gutenberg.net, would be most welcome, and will help to distribute our free electronic texts.
This, I think, does not reflect "common usage". There are millions of deep links into the PG site, directly to the files or into the old (Pietro's) search page.
I disagree. Common usage at content sites, such as online newspapers, is for links directly to content to work for awhile, then to stop working as part of the document management process. But this is not an important disagreement, because I think we will agree on the rest:
Many of them are found inside newsgroups, blogs and reviews where changing them is impossible. I am trying to keep those old links working with redirects.
We should (and I have done so for some time) encourage links to the bibrec page in the canonical www.gutenberg.net/etext/12345 form instead of to the files.
Agreed. But the bibrec page has only existed for a few months. Writing a linking policy that says to link to the canonical etext/xxx location would help.
We also have a canonical www.gutenberg.net/author/Mark_Twain url that gives you all books by that author. This url is used on the wikipedia pages to get a current list of books vs. the often stale and uncomplete edited-by-hand lists. (I myself put in many of them.)
I didn't know about this one. Another reason to write a linking policy (you are guessing who I will nominate to write it, yes?).
Any user that gets on one of those pages can easily navigate to the root page, so it is not necessary to require deep linkers to also set a link to the root.
Yes.
Deep linking to the files, while harmless, is less effective than linking to the bibrec page. The user may not be aware that there are more formats to choose from, he may not be aware that there are newer versions and she may not see the huge amount of other material we have.
Deep linking to the bibrec is harmless. Deep linking to a particular eBook file (especially in the etext?? dirs) is perilous.
Bottom line:
- we should allow deep links to /etext/12345 and /author/Mark_Twain
Yes.
- we should discourage deep links to the files
Yes. That is what the earlier policy I sent was talking about; we didn't have the bibrec pages when I wrote it.
And further on, we should get our subject cataloging up to date so we can offer an url like /subject/Mystery.
Absolutely. If this all seems agreeable to people, we just need to write a "linking HOWTO" or somesuch. As others have said, we *DO* want people to link to us. But we *DON'T* believe people need to ask permission to do so (and I have on several occasions refused to fill out stupid forms some organizations send, asking permission to link to us). We *DO* want people to link to the bibrec pages, author pages, etc. But we *DON'T* want people linking directly to eBook files, because their locations can change. (This is less an issue with the post-10K directory structure, which is far more regular....but since we still have tens of thousands of files in the old directory structure, we should just discourage linking to files directly.) BTW, once we get some sort of conversion on the fly going, I expect URLs like this: http://gutenberg.net/etext/1234/1234.txt http://gutenberg.net/etext/1234/1234.pdf http://gutenberg.net/etext/1234/1234.mp3?maxsize=10m&speed=2x http://gutenberg.net/etext/1234/1234.htm?css=blueplaid&font=verdana and the like...where files we already have are delivered, but files we don't have are generated. I don't think I'd recommend people link directly to such URLs because (a) they depend on converters that might not always be available, or might require additional syntax, and (b) because such a URL is fine for individual use, but if you tell your friend such a URL then their choices for alternate formats are less likely to be evident to them. Further thoughts? -- Greg