But that’s a centralized vs distributed argument. I understand that. By why exactly does a centralized development process have to be orders of magnitude slower than a distributed one? I just think the tooling has let us down -> big time.
Regards, Greg Dr Greg Low 1300SQLSQL (1300 775 775) office | +61 419201410 mobile│ +61 3 8676 4913 fax SQL Down Under | Web: www.sqldownunder.com<http://www.sqldownunder.com/> | http://greglow.me<http://greglow.me/> From: ozdotnet-boun...@ozdotnet.com [mailto:ozdotnet-boun...@ozdotnet.com] On Behalf Of Ken Schaefer Sent: Tuesday, 22 November 2016 2:29 PM To: ozDotNet <ozdotnet@ozdotnet.com> Subject: RE: [OT] node.js and express From: ozdotnet-boun...@ozdotnet.com<mailto:ozdotnet-boun...@ozdotnet.com> [mailto:ozdotnet-boun...@ozdotnet.com] On Behalf Of Nathan Schultz Sent: Tuesday, 22 November 2016 1:53 PM To: ozDotNet <ozdotnet@ozdotnet.com<mailto:ozdotnet@ozdotnet.com>> Subject: Re: [OT] node.js and express But many of the same problems persist on the web, and the web has brought entirely new challenges. The web has all the same issues of DLL hell - different components need different versions of the same component - like the other day I saw a project where there required multiple versions of JQuery due to different component requirements - and the hacks required to ensure that the right version is used for the right components isn't pretty. And apps are no longer isolated within the company - you're not developing for a particular SOE anymore - but rather a polyglot of devices with different operation systems, software, features, sizes, resolutions and capabilities. Testing if anything is a longer process than ever before. Licensing for some software is on a per-user basis on the web, which brings it's own challenges for anonymous systems. Deployment must still go through a full change management process (there are still a multitude of things that can go wrong), and when you consider the same thing could occur today using one-click deployment, sand-boxed applications, and Docker containers, the web doesn't have an ace up its sleeve there either. In my experience, part of the reason that enterprises prefer stuffing all that complexity resolution into the development effort (or getting your vendor to take the hit in their development effort) is that it’s project cost. It makes it much clearer what the true cost of “application X” is to the business. If the alternative is passing the cost to BAU/Ops (in terms of managing interoperability and keeping an end-user fleet of devices running), it becomes far more murky as to what is causing the complexity and how much it’s costing. Whoever is funding the project will cut whatever they can (whether it be security, sociability testing, monitoring instrumentation), and make it Operations problem. As for “click once” etc. that’s just solving a small technological piece of a puzzle. How do I do deployment accounting/licensing etc. via click-once? My current org is ~40,000 users spread from Sydney to Woomera – do not underestimate the complexity of deploying or upgrading anything critical in that type of environment: when we used to run a distributed Active Directory environment (so that local branches could keep running if the WAN was down), simply upgrading AD schema was a 9+ month project, where we ended up auditing every DC (around 150 branch ones at the time) to verify that their out-of-band management cards were working (about 20 needed replacing, or were not cabled), and had to roster tens of support techs to be ready to drive/fly out to a site, just in case we had to pull the upgrade process due to something going catastrophically wrong. The change had to be done over a long weekend, because that was the only time that gave us enough lead time to do an authoritative restore. Now, upgrading AD isn’t particularly important to a bank – and if I was IT leadership I’d be asking: “why can’t I deploy a core systems upgrade, or upgrade online banking during this key window? Why am I upgrading ‘AD’, whatever the f*ck that is” Now, it’s all sitting in our data centres, and we could probably do a scheme change overnight if we had too. There are lots of Ops benefits to centralising all your core logic and systems. Cheers Ken
