670 karma · joined April 15, 2014
I remember when we were to be shut down by AWS because the bill was 2 months due and the administrative contact was a inbox nobody looked at. We discovered it by chance, doing a nightly maintenance: services started to disappear one after another (including the Redis that powered the login, we weren't able to login into the product for half an hour!). We immediately phoned AWS support and asked, no, begged them to turn them up again. They did and we paid everything the morning after.
Hadn't we been in front of the PCs doing the maintenance and witnessed as it was happening, we might have woken up the day after without an online bank (yes) because the production environment would have ceased to exist...
What I dislike here is that it looks just a fancy packing of already existing technologies borrowed from academia like OpenNMT.
I think I am gonna take this with a grain of salt, meaning as the rant that it is :)
It's valuable to hear some dissident voice from time to time because 100% agreement on trending paradigms always creates the risk for the new generations to jump to the cargo cult ship and accept the common wistom as a dogma, instead of appreciating with own reasonment the current state of suggested practices.
As always, virtue lies in the middle and computing is no exception: I was professionally born in the age of in house racks where every SME would own a handful of servers and sysadmins would plan whether colocating applications or buying new iron that had to be accomodated somewhere in the DC. There have been very good reasons to switch from monoliths/SOAs to microservices and from bare-metal to containers/FaaS, reasons that I personally would not trade back for the bad old days of manual deployments: microservices offered live patching of systems with no downtime and imposed design by contract to ensure verifiability of single cogs inside the big clockworks, instead of having to run the full stack in own laptop to ensure end-to-end functioning.
I think the point of the article is really how bad is to swtich to new paradigms for use cases that are well served with single application, single server monoliths (internal inventory management systems, for example) just to change ship and "do what the others are doing".
Yes, the cloud is conceptually the new mainframe, but usability has also improved a lot: the user experience created agility and the agility created the opportunism that creates the commoditization of infrastructure. Is Heroku a sustainable use case for production? No, but it is super for development. Heroku is feedback loop for product development as smalltalk was for pure development.
So it is probably true that following blindly market trends just for the sake of it is a very bad idea, but the author seems also to be suggesting "Java monoliths or die!" which I don't agree upon.
Conciseness is a valuable property which, for web development, PHP seems to have given up.
>Objectively, compared to other languages i've been working with it is more than OK. Despite it's lack of "style" it is easy to understand, host, tests, diagnose and it is powerfull for web applications. >I've been working on a SaaS API with Symfony now for 2 years, our metrics are good... >It is a good tool, it builds good softwares
Informed? These are opinions, dude. Where are the objective arguments? "It is ok, works for me" is not "being informed"
It made sense to use PHP EXACLTLY in the good bad old days because it was just leaner to bang out applications in PHP than with other languages.
Nowadays, I feel that PHP cake has been largely eaten by Python and Ruby.
Why using PHP with such a level of verbosity?
An increasing number of people that spend their full office hours doing meetings, while devs need silence and solitude, complain that remote work sucks for meetings. Wow. What a bunch of egotistic children.
- Hilbert
I suspect this DevOps is working in loneliness, without enough communication with devs. Devs ought to be educated, it's part of DevOps job.
Ship it, ship it
Ship it, ship it
Merge into master and release it
Add some more features, fix some more bugs
it doesn't matter if it don't compiles
Just ship it (just ship it, whoo!)