4 ms·

The points are solid, and early planning where possible is great. MVP often contains uncertainty, so there are a couple of methods I've used to afford flexibili

The points are solid, and early planning where possible is great. MVP often contains uncertainty, so there are a couple of methods I've used to afford flexibility in the future (although it still helps to codify this better when the flux settles) For applications with low expected item limits within taxonomies it can help to use a numeric id, combined with an open-ended rewrite of the user-readable content. http://www.example.org/items/12-really-awesome-thing http://www.example.org/items/12-really-awesome-thing RewriteRule ^/items/(\d+)[a-zA-Z0-9_\ -]+$ index.code?route=items&id=$1 Although you probably index your titles anyway, simple numeric key lookup makes the DB layer happy. Combine this with rel=canonical to your preferred url, and you can be more flexible with case, specific terms etc. It has the added bonus of reducing 404s through typo'd links and such. A couple of other anecdotal points: * Underscore is considered a letter, not a space, for most purposes * Pages should either have a (faux) file extension, or no trailing slash. Reserve trailing slash for directories * Rel Canonical also allows you to carry non-critical data in query string using [QSA] in the RewriteRule, without polluting the search index with duplicate content. Consider this for, example, default view of output (CD cover vs. title list)

minouye 16y ago
Interesting point on embedding numeric IDs. Looking at the Backcountry.com example, have you ever attempted to place all content one level beneath the domain (i.e. all content has a directory depth of 1)? - Category: http://www.backcountry.com/mens-clothing http://www.backcountry.com/mens-clothing - Item: http://www.backcountry.com/smith-phenom-goggle http://www.backcountry.com/smith-phenom-goggle - Brand: http://www.backcountry.com/smith http://www.backcountry.com/smith I was rather baffled by this approach, but it seems like it was a rather recent change and I'd assume it's been well thought out. Intermingling brands, categories, and products makes me think that they must use a lookup table of sorts (I don't know how you'd use rewrite rules), but I've never seen an implementation like it before. Any idea on how to approach implementing varied content all on the same directory depth?
- tomwalsham 16y ago
  For me, a low-value numeric id has never been an impediment. Aesthetically you might lose a small amount, but many of us use surrogate indices in our databases and it fits well with that design pattern. The backcountry example smacks of overzealous in-house SEO to me - the observation that Google (in particular) tend to favour the lower level of structure could lead to choosing that output, but ignores that the first three levels appear to hold comparable weights. I'm not convinced it would offer much benefit. Matt Cutts has indicated in the past that 'natural' levels make sense, maybe 3-4. First level entries do have use (see the 'enhanced' results for some sites based on sitemap priorities), but having 100s of them will likely remove this useful trigger. From a technical standpoint it's definitely creating obstacles. You create an additional layer of restrictions in the database index. They must have a table of unique level-1 indices. So if there is a brand that corresponds to a generic term (not great for trademarks, but it happens), you will get a uniqueness collision. It seems well executed, and no doubt their CMS input will inform them of clashes, but for me this is genuinely not as useful as a 'breadcrumb' structure from a UX perspective. They've essentially introduced complexity for the sake of an assumed SEO advantage that my anecdotal evidence can't support, and in doing so have lowered the utility of their urls, per the principles (of which I agree) of the parent article. That they're buying the term "smith phenom goggle" from adwords, and are stuck on page 2 of the SERPs might back this up.

minouye 16y ago

Interesting point on embedding numeric IDs. Looking at the Backcountry.com example, have you ever attempted to place all content one level beneath the domain (i.e. all content has a directory depth of 1)? - Category: http://www.backcountry.com/mens-clothing http://www.backcountry.com/mens-clothing - Item: http://www.backcountry.com/smith-phenom-goggle http://www.backcountry.com/smith-phenom-goggle - Brand: http://www.backcountry.com/smith http://www.backcountry.com/smith I was rather baffled by this approach, but it seems like it was a rather recent change and I'd assume it's been well thought out. Intermingling brands, categories, and products makes me think that they must use a lookup table of sorts (I don't know how you'd use rewrite rules), but I've never seen an implementation like it before. Any idea on how to approach implementing varied content all on the same directory depth?

tomwalsham 16y ago
For me, a low-value numeric id has never been an impediment. Aesthetically you might lose a small amount, but many of us use surrogate indices in our databases and it fits well with that design pattern. The backcountry example smacks of overzealous in-house SEO to me - the observation that Google (in particular) tend to favour the lower level of structure could lead to choosing that output, but ignores that the first three levels appear to hold comparable weights. I'm not convinced it would offer much benefit. Matt Cutts has indicated in the past that 'natural' levels make sense, maybe 3-4. First level entries do have use (see the 'enhanced' results for some sites based on sitemap priorities), but having 100s of them will likely remove this useful trigger. From a technical standpoint it's definitely creating obstacles. You create an additional layer of restrictions in the database index. They must have a table of unique level-1 indices. So if there is a brand that corresponds to a generic term (not great for trademarks, but it happens), you will get a uniqueness collision. It seems well executed, and no doubt their CMS input will inform them of clashes, but for me this is genuinely not as useful as a 'breadcrumb' structure from a UX perspective. They've essentially introduced complexity for the sake of an assumed SEO advantage that my anecdotal evidence can't support, and in doing so have lowered the utility of their urls, per the principles (of which I agree) of the parent article. That they're buying the term "smith phenom goggle" from adwords, and are stuck on page 2 of the SERPs might back this up.