Sorry about that. Didn't mean to be a jerk if that is what I did, but in my own defense, I said "trust me on this one."

nah you weren't being a jerk, just mildly frustrating. i still you.

If that is what my writing looks like when I'm tired, you can only imagine the horror that is my code at moments like this.

To be fair, I just used Amazon as an extreme example. I would argue even much smaller scales should be saving all of their data (even versions of product descriptions) and I suspect we disagree at where that line is. But I can live with agreeing to disagree.

So I finally made my own website and am hosting it off an old computer out of my bedroom. It's very basic at the moment, but it feels pretty damn good to have it up and running and know that you did everything on your own from beginning to end.

Main issue I'm having is I had to reroute port 80 because my ISP blocks it. As a result I had to enable framing so the url doesn't convert to my IP. All is well except the site is no longer rendering correctly on mobile. Any experience?

Are you talking about Amazon the retailer or an Amazon Marketplace third-party seller? Amazon is not eBay - third-party sellers are there to provide the long tail and to help optimize their internal services using market prices but they are secondary to their main business of just selling stuff.

I think everyone would agree with your point for some small seller on Amazon Marketplace - but the point of bringing up Amazon, I think, was not to talk about the some small seller on Amazon Marketplace, but Amazon the biggest online store.

Passwords might not be a time where you want it tracked but could have value somewhere to someone if you had the data handy.

You could be like... users between the age 20-30 who are female change their passwords 8.35% more often than anyone else by using the password reset form. So maybe you could do something with your UI for that subset of users to help them pick more meaningful passwords.

Or maybe you could narrow down a tight range of users who regularly change their password manually. Over time and with other data points you could probably come to a conclusion that this user cares about privacy and is knowledgeable when it comes to technology, so maybe you can offer him certain types of products.

Of course this won't be applicable to everyone but that's a very reasonable use case for wanting to track when someone changes their password and the mechanism they used to reset it.

Also I would want to version all content/copy honestly. It's so nice being able to just roll back to a previous version, especially if you're at the point where A/B testing is meaningful.

I'd probably track something like password reset frequency through an event-based analytics framework at the client level.

Easy roll back is a great use case for versioning. Keeping an audit trail of changes made to the site is another one I thought of too.

Yeah having a trail of changes for code is super valuable but I think people under estimate how valuable it is for content too.

A lot of time you have your content stored in a DB and once you change it, that's it. It's gone. You'll likely have backups happening on a regular basis but sometimes you just want to compare the version from a few days ago to today just to take a quick glance, happens to me a lot.

Importing the sql dump locally just to do that is such a huge hassle. It's like using zip files and FTP to backup your code.

Like what? Convert the login button to rose.gif?

This is one of those areas where I feel that you'd have a hard time striking a balance between asking for too much info and just finding a decent baseline. If a site is asking for my gender, address, name, and other non-essential information, then there better be a compelling reason.

I always liked this article by MailChimp in regards to password retrieval. I know the best-practices security counsel will have a heart-attack, but the data is quite surprising:

https://blog.mailchimp.com/social-lo...rent-worth-it/

If you are doing A/B testing, wouldn't you have two versions anyways?

If you want to keep all of your history, you can easily create a persistent schema by defaulting all article insertions with current_date and pulling up the current articles with max(date). No need for 800 manual backups.

There's a silly program we use at work. We always know when they "update" their software: since every new push was not well-tested, we run into a million odd-ball bugs. Even without new updates, I submit at least one bug-report a week to the company. Today, there were three of them submitted, but thankfully none by me. For kicks, we sometimes like to play "reverse engineer the bug" or more accurately, we take stabs at how the **** "this one" happened.

This one is from a few months ago and it is the one that bothers me the most because I have one suspicion about how it happened, but I honestly don't believe my own theory.

A coworker was getting an error: "There is already a column called 2.25. This is a contrived number, but it was reflected in the price.

My coworker had a spreadsheet, and it looked like this, transposed, of course:

product id: 12345
product name: do do bird
price: 2.25
price plus shipping: 2.25

I quickly recognized the issue. For some reason, the program was skipping the first line of the file and reading in the second line of the file as the headers. I don't him to copy down the header line and lo and behold, the upload was a success, sort of. Since he had missing data on the top row of the product id and product name columns, the system read it as nil and didn't feed any of the product ids in. We had the pricing and other information, but no product names to go with it. This was a clerical mistake. He updated and everything worked like a charm.

This got me thinking about the internals. Aside from not throwing an error for missing vital information, how is it reading the top line of the file and using that? I suppose there is some need be flexible on the order of the headers, but are they doing this:

-- feed in the headers
-- loop -> concatenate the headers into the following clause:

insert into myTable (header1, header2, header3, ....)

??? What am I missing here??? And how good / bad is this if they are?

100% in the track all create, modify etc. with timestamp and user if possible camp. Never delete, change state of stuff to removed or something instead.
Well I worked in ERP where you pretty much have to do it this way (financial and other regulations require audit trails)

davet,

In my hypothetical I was thinking of a match making site where gender plays an importance.

Also your current_date idea is tracking everything, it's still an audit trail. The implementation details don't matter.

I don't know why I always forget to use underscore.js... so many useful functions for 5kb.

Nchabazam,

There's lodash too. http://lodash.com/

Cool, thanks.

So I finally got around to asking the head of our (giant company) QA department what their standards are for testing legacy IE (we support IE8+) and her response was:

"We only test legacy IE on request, otherwise we stick with the current version: IE9"

Serious question... is there a user base for IE9? It seems like it fixes some of the annoyances of IE8, but still doesn't have many users. I'd prefer to ignore that as well and build cool stuff with like websockets.

I thought Microsoft was pushing IE10 now but not sure. IE9????

I think windows 7 shipped with IE9 then upgraded to 10? 10 is fine and modern, 11 is decent, and 9 I never hear about. I know it can handle some newer stuff, but isn't 100% compatible with things like websockets. I just don't think many people would be on it.

I know the rose.gif was a snark, but I wanted to point out that you can collect massive amounts of data but have no conclusions about what to do with it.

These people are the type who do not click "keep me logged in," don't click "Let [your browser] save your password," or use other tools to make their lives easier and many of these tools are smack-center for them to use. Heck, they practically expect the sites to keep them logged in.

In general, I would say that more upfront work and following the trends of successful sites would be a much better time investment than setting up your site to collect this sort of data.

But it is better than combining and comparing a bunch of manual backups.

All of that from yesterday got me thinking of it at work. I decided to comb some of the reasonably large-selling products and check out the conversion rates. A somewhat surprising find was that a large portion of the top converters have 0 lines of product description.

Obviously, you shouldn't design a crazy-ass site and hope data will point you in a brand new direction for a relatively solved problem. But collecting data about your users is super valuable and should be a pretty core part of any tech business regardless of scale (assuming you want to attract new users/customers and grow).

dave,

What types of products have the best conversions? I also think having some data on the user would be interesting in this case.

For example knowing how much time they spent on the site and how many pages they browsed during the session they made the purchase is really valuable. It's knowing for sure if the guy just popped in to buy something because he knew he wanted it vs browsing 15 product pages and shopped around.

Of course there might be impossible to track things that skew the data like the guy went AFK on the product page for 20min but I think overall having the data would be way better than not having it.

What's the word on Facebook's Hack language?