Hacker News Reader

Comment by mgkimsal | Hacker News Reader

It's pretty much the same in software too. I often get requests for functionality and when I say "no, we can't do that", I get "Why not? Amazon does it".

Hrmm... Amazon has dozens of people working and supporting just that one feature. You're trying to engage me to do an entire project. On a fixed budget. With a fixed time frame. And you've changed your mind 3 times in the last 3 weeks on key points.

Of course, yes, there are amazing things you can do with software that weren't remotely feasible even just 5 years ago. But there's always a moving target - the market leaders (Apple, Google, Amazon, etc) are constantly pushing the boundaries of what's considered 'normal', and most people have 0 idea of the real cost and effort involved in having the functionality come across as polished and error free as the big boys.

View on HN

On the other hand, doing even simple things at Amazon scale is WAY more difficult and time consuming than doing them at practically zero scale. There are plenty of examples of features that can be accomplished using quick and dirty methods that Amazon couldn't dream of using.

Could you give an example something Amazon would have a challenge doing due to its scale?

It's a shame someone down-voted you. I'm not sure if you were being rhetorical, but this question is easily answered because the answer is, quite simply, "just about anything".

Want to add anything new to the page? Well, at Amazon-scale, your little widget is going to have to go through some pretty intense design review and load testing, because at over 150 million page views per day, tiny time penalties can add up to significant capacity issues.

Amazon is also notorious for deep introspection of user behavior, so any little thing you want to add is going to have to go in to the big bin of factors that might influence a customer to "click the button" (or not). Data about your new feature will have to be integrated in to the models that are used to evaluate user behavior, or at a minimum, the group who handles that will have to sign off on not tracking the impact of your new feature on user behavior.

The whole concept of "the group that manages that" isn't unique to Amazon, but once you pass a certain size, you don't simply add features or new products without a involving lots of people. That is a challenge unique to Amazon-sized companies.

Reminds me of the many times someone tells me "oh you're developping apps ? I've got a great idea : ". 9 times out of 10 this idea involves rebuilding google search engine and facebook, plus a third thing on top.

http://howismyappidea.com

Damn, I seriously expected something like a vote system

A voting system wouldn't get it right in 99.9% of the cases.

Why can I not give you a bounty here like on Stack Overflow?

Instead of "no, we can't do that" give them the appropriate quote.

I was going to say the same thing; this actually works really well. "But Amazon does it!" becomes a less compelling argument to people when they see how much they'd have to spend to be like Amazon.

Nah, this doesn't work. Clients think you're just lying and trying to take advantage of them.

If you have a client asking for million dollar features on a 100k budget, the only solution is to find new clients. Some people are beyond help.

This is why building trust by starting small and delivering exactly what you said you would for exactly the price you said is key.

Clients and stakeholders alike will never cease the asking for more than you can reasonably deliver because they have absolutely no way of knowing how reasonable their request is. That is why pulling all nighters and heroics are so bad for developer / business relationships. They set a precedent for what is reasonable or possible.

If the client or stakeholder never really can have the data necessary to fairly judge whether you are "holding out" or not then the only thing you can do is build personal trust by doing what you say you will, not a damn thing more, not a damn thing less.

I generally do if they press the point, or some similar variation. "Oh, but Google does XYZ". And again, yes, but... their budget has 3 extra zeroes on it. etc...

Exactly.

Hardware, software, anything really - there's probably an analog to this in most things.

Personally, what I run up against most often aren't questions about fulfilling the initial requests, but the aftermath of someone focusing on specific details to the exclusion of the bigger picture.

An organization spending a few million on software licenses, but nothing on staff / support being the most common example.

I feel like people in tech are particularly prone to this for whatever reason.

I'd wager a person will typically to sustain more criticism for a sub-optimal, but complete, functional implementation than a disjoint pile of fine-tuned components that doesn't add up to anything.

> Of course, yes, there are amazing things you can do with software that weren't remotely feasible even just 5 years ago

Can you provide an example to back up your claim?

Software defined radio[1]

[1] http://en.wikipedia.org/wiki/Software-defined_radio

Bad example; SDR existed in prototype form as early as the 80s, and the USRP has been commercially available since 2005.

Yes, but today you can order a USB SDR radio for under $10 and you have a crapton of FOSS software to use with it.

I would personally point to a framework like Meteor as my example. Having a framework handle all of the reactive parts of your code with no additional work, and giving the end user a way to simulate database updates on the client side while it's being processed server side is a huge win for performance.

Now of course none of this was impossible to do before, but the advent of all major browsers now supporting web sockets, and the existence of this framework and community which has already solved all of the hard parts of the implementation makes it actually feasible.

I was thinking of things like deploying auto-scaling SOLR clusters for hundreds of dollars, vs tens of thousands for dedicated search infrastructure years ago - maybe a bit more than 5 years, but that's the gist.

Twilio. Maybe Twilio is within the 5 year window too.

Mobile apps in general - the hardware infrastructure in peoples' pockets lets us do stuff with software that was not on most peoples' radars 5 years ago.

Reply on news.ycombinator.com