Disqus - Latest Comments for rubypond

Re: Heroku is no longer the hobbyist's friend

Glenn Gillen — Wed, 21 Oct 2015 18:12:55 -0000

So you get free SSL, at https://your-app-name.herok...

What you're asking for is to have SSL on your own branded domain. Is it really a "hobby" at that point?

Re: Heroku is no longer the hobbyist's friend

Glenn Gillen — Wed, 21 Oct 2015 18:10:37 -0000

Yeah, I think people often make the mistake of comparing Heroku costs (which include some amounts of automation/fail-over/backups, especially on DBs) with just the raw infrastructure costs elsewhere. They then assume their own time is worth nothing, that they won't have to spend a single minute doing any of this stuff themselves, or that it magically takes care of itself.

SSL is definitely a sting here. Kinda ironically though the ExpeditedSSL integration is one of the things (along with Postgres) that keeps some of my apps there. $45/mo to never have to worry about a a certificate accidentally expiring, along with quarterly cert rotation and the someone else making sure it's up-to-date with the latest best practices is a simple decision imo. I've only so many hours in the day to keep track of things.

Re: Heroku is no longer the hobbyist's friend

Glenn Gillen — Wed, 21 Oct 2015 07:32:17 -0000

On the SSL point, the reason for the cost is partly to do with the broken state of TLS/SSL and using it in a multi-tenant environment. When you add it you might have noticed you get a new CNAME to point your DNS to. That's because a whole new dedicated Amazon ELB is spun up just for you, so that you can have your SSL certificate installed on it and it can then manage SSL termination for you. Last time I did the math on that I think it cost somewhere between $18 and $19 a month (depending on the number of hours/days in the month). Add in bandwidth costs and I suspect Heroku is making a loss on them, before you factor in their own ops and support overhead for providing them. I'm still unhappy that it costs so much, but I also understand why it isn't free. They do actually give you a free option, just that it's on a Heroku branded domain.

As for the Cloudflare type approach, that creates a whole world of different problems. If you find your SSL terminating on the same instances as some questionable neighbor you might find a number of people can't access your site at all. Lots of jurisdictions, or just over-zealous ISPs, actively filter the content the users can access. And they often do it by IP address, which is how your site gets caught in the mix. While I lived in the UK I think that included every mobile network, and it's a lot of ISPs back here.

We disagree on postgres too :) Spinning up a database is just the tip of the iceberg. The real cost comes from preventing things from going wrong, and resolving them if they do. I'm happy to pay $50/mo for automated backups, multi-site storage of write-ahead-logs, automated monitoring, automated recover/failover if an instance goes down, rollbacks, ability to fork/follow, and dataclips. Setting up and maintaining all of that for myself would easily take more than 30min/month. So it's way cheaper to let Heroku do it.

I completely agree with everything else though. Chat bots have to move. Other little trivial things too. I'd been meaning to play more with CoreOS and Kubernetes for another app I had that couldn't run on Heroku. Between that or Convox.io I suspect I'll have a single platform to run all of my apps, big or small

Re: I quit the techÂ industry

Glenn Gillen — Wed, 10 Jun 2015 08:42:15 -0000

Great and timely perspective. Today I've just ticked over 12 months since I left my tech job, most of my time since has been spent with my family and enjoying watching our son grow. It's really helped me come to terms with what's important and how (in)significant most of what we do in a given work day really is, which makes the anguish it often causes all the more pointless. I'm going to start my own business soon and finding a way to provide true work/life balance is going to be one of the most important things for me to instill culturally. Not the lip-service fake balance I've seen so many tech companies where what really happens is the lines between work and non-work just become blurred. I wish you all the best for whatever comes next and hope that the period in between is as restorative for you as it has been for me.

Re: Add-ons for Production Apps

Glenn Gillen — Thu, 18 Jul 2013 02:27:07 -0000

I can second what Daniel has said. As Heroku Postgres has your hot data in RAM it's pretty fast, and has JSON support in the form of hstore. A big benefit of a dedicated caching solution is the automatic expiry so you don't need a separate process to go through and do the cleanup.

On the addons.heroku.com app we're running a cache in front to store the rendered output of the most frequently accessed pages. In many cases it can avoid a DB hit completely, and we get some failover by having two accessible proxies for the cache. If that fails or hits a timeout, then the page is still rendered just a bit more slow from n queries in the database.

If it's possible to easily replicate your use case I'd consider deploying an example to a staging app and then running a load test against different implementations using something like Blitz.io (https://addons.heroku.com/b... to know which one works best for you.

Re: Why You Need a Git Pre-Commit Hook and Why Most Are Wrong

Glenn Gillen — Wed, 16 Nov 2011 14:33:13 -0000

Great post. I think part of the reason why a lot of pre-commit hook examples check the whole project is because there is no way to enforce certain hooks as mandatory across the team. So in the example above JShint would only be run on the files I touched.

At least that's why I do it. Trust no one! ;)

Re: 13 Things You Must Do Every Week As A Startup CEO

Glenn Gillen — Tue, 05 Apr 2011 13:08:43 -0000

My only criticism, why should this advice be limited to "startup" CEOs? I wish every CEO I'd encountered considered this every day.

Re: 37signals vs Zappos

Glenn Gillen — Fri, 11 Feb 2011 12:13:00 -0000

I spent the first half of my career setting up contact centres in numerous countries, the latter half working in something more like start-up environments. So hopefully I've got some perspective on the reason for the differences between both.

Firstly Zappos. It's no doubt a great company to work for, the contact centres I'd setup were regularly voted the best places to work which is highly usual for that sector. They're usually drab and lifeless places, where people are essentially chained to their desk by a headset. Toilet breaks often need to be cleared with a manager first, you need to literally clock in and out each day, and to top it all off most of your time is spent trying to calm down angry people who are calling to complain about something. The staff really are fighting on the front line and they have a hell of a task to stay happy and motivated about what they do because at the end of the day, it's a pretty sucky job. The solution to that? Make everything else fun. The workplace and the surroundings have to work as a distraction from the other negative impacts on their job. Some autonomy to decorate their environment creates an outlet for expression in a role that is otherwise heavily regimented and scripted.

Compare that to the average life of a developer, most of us actively love the work we do. Writing code, producing something people get to use, and the sense of accomplishment that comes with it. Jason is on record as saying (paraphrasing here) that interruptions are rude, you're basically saying whatever I want is more important than whatever you're doing. Mental state and flow is important to a developer, distractions are preferable kept to a minimum, and that includes the environmental ones. It's no surprise that their office reflects that, and I think it does so for good reason.

I think it's just two companies servicing everyones innate desire to be just a little bit creative in the best way they can. For one it's in the environment they create, the other it's the work they produce.

Re: 37signals vs Zappos

Glenn Gillen — Fri, 11 Feb 2011 07:13:11 -0000

Re: Story Point Algebra

Glenn Gillen — Thu, 10 Feb 2011 03:35:44 -0000

I've always found the Fib estimations a bit weird for precisely that reason. Not because it's not a good way of estimating complexity (I'm indifferent on that) but because, as you intimate, when you stand back objectively ask "what are we trying to do here?" you're wanting to know how long it will take to complete and how much work you can schedule.

The places I've worked where the points reflected this the most consistently have been where estimation was done in days. 0.5 day, 1 day, 2 days, 2.5 days at the most. Anything beyond that and you've probably not broken the story down into the simplest thing to get the job done (improvements can be a later story for a subsequent iteration). Then just map that effort to some arbitrary points system (1, 2, 4, 5) so that you remove yourself from explaining load factor to clients, they just get to schedule a set number of points each iteration.

Re: Cucumber vs. Steak

Glenn Gillen — Tue, 01 Feb 2011 04:45:04 -0000

It *could* help people separate problem/solution, but most don't. It's just as easy to do it wrong with Cucumber and as you've said the example web steps actually lead you the wrong way and as a result many people do it wrong.

It takes conscious effort from the developer to separate the two issues, and at that point you could expend the conscious effort in making sure your tests in any framework are readable and testing the right things.

And so I think you've succinctly identified my main issue with Cucumber, and it's not really with Cucumber at all. It's that people seem to think using it and saying "BDD" magically means you've fixed what was essentially a process or communication problem. You've still got that problem unless you've changed your approach, you're just using a different tool to keep doing it wrong.

Personally I stick with index cards for the plain English stuff, it's easier for me to shove them in front of the client and go through them that to convince them to look at the output of a test run somewhere. They're happy to take my word on whether the card is "green".

EDIT: Forgot to mention, I thought it was a great summary of the real pros/cons (as opposed to the usual feature comparisons). I'm not convinced Steak is any better at solving this problem.

Re: Flash Ad: BMW Burns Logo onto Cinema-Goers’ Eyes

Glenn Gillen — Thu, 16 Dec 2010 04:15:12 -0000

Living in Europe (does Australia qualify as "Asia" these days?) and going to the cinema on average once a week I can anecdotally report that mobile usage once inside is incredibly low, maybe one person every 3-4 visits but only ever during the non-Movie advertisements even then.

High mobile usage in a country does not automatically imply a decline in decency standards. Cinemas, theatres, etc. are typically places where actions that distract other patrons are frowned upon. I'd be saddened if that changed.

Drawing a false correlation from data doesn't protect someone from also looking arrogant and naive.

Re: Postmodern - Mining RubyGems from Ore

Glenn Gillen — Tue, 14 Dec 2010 07:07:28 -0000

A bit late to this conversation, stumbled this way from a post postmodern made about Ore on the ruby learning blog.

I agree with Ryan, the main problem is rubygems should be doing most of what these tools do. Until that day I like the look of Ore, certainly seems like one of the better gem-cutting tools so thanks for pointing it out. I'll definitely take a deeper look.

Re: PayPal+VP+On+Blocking+WikiLeaks%3A+State+Department+Said+It+Was%26nbsp%3BIllegal

Glenn Gillen — Wed, 08 Dec 2010 12:01:55 -0000

Account closed, feels good to be done with this horrible company.

Re: Should Nesta be a gem? - Nesta, a Ruby CMS

Glenn Gillen — Wed, 08 Dec 2010 09:07:58 -0000

As you said, would make mounting it in Rails much easier than it is at the moment. Plus it would reduce a lot of duplication on my server as I have a nesta checkout for each site I host, I could just have 1 system gem instead.

Re: Running background jobs with resque - /dev/mull

Glenn Gillen — Mon, 08 Nov 2010 05:48:29 -0000

You've been giving examples using MetricsList but the ResqueJob implementation is for MetricsUnsavedList, is that by accident or is it indeed a different class to the one that has perform/while_updating_job_status?

The only way you can return the class name easier is to drop the self., as it's implicit given you're in a class method. On that note, the @job assignment is assigning to a class variable given the scope you're in. Running self.while_updating_job_status twice in the same ruby process will have the second call replace the assignment from the first which is almost certainly not what you want as it will create a right mess and update the incorrect records should two jobs be checked concurrently.

Re: Our Git Workflow: Private Development, Public Releases - Braintree

Glenn Gillen — Tue, 17 Aug 2010 16:00:53 -0000

I have to agree with Travis and Henrik here, and I don't buy the "feel that topic branches generate too much clutter". Branching in git is so quick and trivial that I'm always a little confused when people don't take advantage of it.

Given that within git a branch, tag, and pretty much anything else is ultimately just a SHA pointer to a commit you may as well just do your development on master, rebase --interactive to squash your commits, and then tag whatever you release when you push to origin. Your 3 branches are just pointers to a specific commit SHA, 2 of them to the same SHA that you tagged, and one of those is only there so you can do a squash (which you can do with rebase).

I'm not advocating that approach though, I prefer to see a full commit history so if there is a particular change I'd like to pull in myself for whatever reason it's much easier to cherry-pick rather than delving through a massive diff.

Re: 9 Ruby on Rails Backup Solutions

Glenn Gillen — Tue, 23 Mar 2010 13:05:39 -0000

I've got one that I rolled myself, I've been meaning to package it up as a gem and make it public. Will try and do that in the next couple of days.

Watch http://rubypond.com/blog/de... for a post on it

Re: Nesta on Heroku - VorpalCode

Glenn Gillen — Fri, 05 Feb 2010 04:04:01 -0000

I've been using nesta too lately and it's been great; I've also opted for keeping the content in the same git repo. As a result, each nesta site is now just a separate branch off master so I can pull changes from Graham's github upstream and then easily merge them into each site when I need to. It's made managing multiple sites really simple, I get the benefit of a consistent CMS/framework without having to restrict every site to follow the same structure entirely.

Re: JRuby, and why bouncers are assholes | Ruby Pond

Glenn Gillen — Fri, 29 Jan 2010 04:26:24 -0000

one more test

Re: JRuby, and why bouncers are assholes | Ruby Pond

Glenn Gillen — Fri, 29 Jan 2010 04:26:10 -0000

testing another comment

Re: Rails Underground - Day 1 | Ruby Pond

Glenn Gillen — Fri, 29 Jan 2010 04:25:42 -0000

testing comments

Re: Using rspec have_tag

Glenn Gillen — Sat, 16 Jan 2010 05:42:26 -0000

Seems fine now, there's an ongoing problem with the code formatting lib I'm using in Radiant, I'm in the process of porting over to something else.

I love the idea of forcing a login to see the code though ;)

Glenn

Re: scRUBYt! Gets Plugins! | Ruby Pond

Glenn Gillen — Thu, 08 Oct 2009 01:51:47 -0000

Rup,

We've not made an official switch to skimr becoming the main and
stable branch purely because we don't have a viable replacement for
scraping sites with AJAX (the current Firewatir integration is a bit
gnarly and not as reliable as we'd like).

I'll be looking to integrate with celerity to allow us to to headless
javascript scraping, once that's done skimr will finally become the
production release and available as a gem. Until then you need to
clone the git repo, switch to the skimr branch, and then include the
checkout in your file:

git clone git://github.com/scrubber/s...
cd scrubyt
git fetch
git checkout -b skimr origin/skimr

Then in your project:

base_path = "/path/to/checkout"
require "#{base_path}/lib/scrubyt"

Hope that helps. If not let me know,

Glenn

2009/10/7 Disqus <>:

Re: Web Spidering and Data Extraction with scRUBYt! | Ruby Pond

Glenn Gillen — Thu, 08 Oct 2009 01:44:32 -0000

Skimr has different dependencies than the current stable release of
scrubyt. It actually requires far fewer external libs and just one
different one, Nokogiri instead of Hpricot.

I've a shiny new MBP arriving this week and I'm going to try get away
with ONLY running ruby 1.9 on it, so any bug will hopefully be solved
shortly. I'll also be tackling full jruby support if we can.

Glenn

2009/10/7 Disqus <>: