Disqus - Latest Comments for amitkoth

Re: How Argo Tunnel engineering uses Argo Tunnel

Amit Kothari — Thu, 27 Aug 2020 13:40:26 -0000

What's the difference between Argo Tunnel and Railgun?

For the purpose of accelerating traffic - the way Argo Tunnel is marketed on your website right now, between origin and the outside web, why the difference between Argo Tunnel and Railgun?

Finally - why can't Argo Smart Routing be used on just traffic to sub-domains instead of being charged for/applicable to - an entire domain?

Re: Rel=prefetch and the Importance of Effective HTTP/2 Prioritisation

Amit Kothari — Mon, 03 Aug 2020 15:41:53 -0000

@andydavies thank you for your response, much appreciated.

On the specific use case of fonts, we thought a much simpler approach would be to take advantage of HTTP/2 multiplexing is simply to make the fonts natively hosted from our own domain. We cache everything on the website for Tallyfy and since Cloudflare natively caches fonts in .woff2 format for example, that does the trick. That saves the round-trips and DNS/TLS resolution time for fonts.googleapis.com and fonts.gstatic.com (both).

Now to the more general case and your specific comments:

1. On whole HTML responses - I absolutely agree with the whole request having to be fetched and non-streaming issue. Dealing with the vagaries of malformed HTML, re-writing behaving badly, etc. are far too clumsy/complex.
2. On pre-connects in head/body - the real gain, I think with preconnects is precedence being set, which I hope - is respected in browsers that support it. Is the notion of precedence a useful property for assets considered critical in CSS/JS vs. just letting the browser discover things quickly? NB - Not setting "crossorigin" correctly on the tags was also a cause of issues for us in the past, dealing with CORS. Chrome doesn't care about pre-connects if certain non-native origins are not set right.

The other important thing to consider is what is defined as performance in general, where to me - Google Web Vitals is far more indicative of "render-ready" performance than just request/response at a micro-level. The macro SEO benefit is via that specific set of measurement, I believe. This means human-readability and readiness should take precedence over deciding what resource is a preconnect or even a preload.

On a separate subject ... re Cloudflare.

The thing (to me) that's most exciting for the performance (of an API) is Cloudflare Workers and Cloudflare KV store.

If you could cache content at the edge on KV store, an API or request can be served at the edge (with persistent data recall in milliseconds) - with no round-trip to the origin at all. There's also no cold-start issues on Cloudflare Workers vs. AWS Lambda functions.

The problem then becomes syncing our master data (on a Postgres database) with the edge-cache on the KV store. But hey - that's another story, for another post :) I have specific ideas for that, perhaps we can do a guest post here or on our own site about the myriad of things we have done.

In addition, because we set our entire website to "Cache Everything" on Cloudflare, the pages are themselves cached at the edge, so paying per-request for Cloudflare Workers gets expensive.

On website performance vs API performance ...

We're far more likely to see benefits from Argo Smart Routing to our API for latency issues via far-away requests to our API origin. REST API's have micro-requests and stateless micro-payloads, where latency matters much more than web page issues like preconnects for many resources on a web-page, imo. A single REST API call is really just a single call, with a single verb, that's limited by the speed of light to serve the call.

Re: The Edge Computing Opportunity: It’s Not What You Think

Amit Kothari — Mon, 27 Jul 2020 14:39:22 -0000

I strongly believe in this - all amazing points, especially the one about compliance, where you essentially guarantee that workers will run local to the country of the user.

However, the biggest problem for us (we are a SaaS platform) is that we are API-driven and all master data has to persist in a database, which is on AWS, at an origin - the source of truth, and of course, the fact that SQL is still very important.

This means we cannot guarantee data is stored in that jurisdiction, because beyond compute - Cloudflare does not have a database that's serverless and edge-driven.

You are getting close with KV store, but you actually need a globally distributed version of Postgres which runs serverless.

Do the database (final resting place) at the local edge too - and you will overturn everything, everywhere. Seriously. Please do it. I'm ready to move.

Re: Introducing Workers Unbound

Amit Kothari — Mon, 27 Jul 2020 14:07:32 -0000

This is awesome, and we're pretty excited at Tallyfy with lots of use cases for Cloudflare Workers/similar.

Beyond compute being unbound at the edge - the biggest problem we have is storage and simple database-like capabilities i.e. the limits/issues of KV store.

After all, it's very difficult to just do compute on it's own and have nothing/no storage requirements whatsoever. In the end, data has to be updated or persist somewhere, such as the origin.

This means a round trip back to origin. So we have two issues to make this "compute at the edge" idea work properly:

1. Compute + fetch data into KV store (so everything at the edge).
2. Sync writes into our master database at origin in the event of a write, but also spread the new values into our KV store, keeping our origin database and KV store "cache" in sync.

With storage and KV store at the edge done properly and consistently, it's far more useful than just compute at the edge. The ultimate thing would be something like a Google Cloud Spanner for the database, but until then - many things are origin-bound.

Re: Serverless Rendering with Cloudflare Workers

Amit Kothari — Fri, 17 Jul 2020 13:10:55 -0000

Thanks for this! Question - if our SPA is an AngularJS app (and we want Google to crawl it normally, for SEO purposes) - how exactly does this HTMLwriter idea work for AngularJS apps?

Re: How To Create a Business Playbook™ (aka SOP’s: Standard Operating Procedures) for Your Startup

Amit Kothari — Wed, 13 Nov 2019 16:50:41 -0000

Thanks @Dan Martell - how would it be tested?

Also - where/who do reviews come from?

In terms of rewards - who gets them?

Re: How To Create a Business Playbook™ (aka SOP’s: Standard Operating Procedures) for Your Startup

Amit Kothari — Sat, 12 Oct 2019 18:18:04 -0000

Awesome article and vid, Dan!

Big question - how important is "doing it" vs. "having it" in terms of playbooks?

We've made https://tallyfy.com specifically to let you make playbooks, SOP's, etc. but focusing real hard on "doing it" too!

Do you also thing passing through/collecting info while you're "doing it" is important? How important is collecting info as you do a playbook in real-time, compared to just following a list of tasks?

We just built a Slack plugin that lets you do things in Slack too :)

Re: Introducing Modern Approvals in Flow

Amit Kothari — Sat, 24 Aug 2019 17:18:27 -0000

If anyone needs a bunch more features, or proper/complex approval workflows - check out Tallyfy - which also has a flow connector here:

https://docs.microsoft.com/...

Re: What I Learned Rescuing Our Startup From Death

Amit Kothari — Fri, 02 Aug 2019 08:05:17 -0000

Maybe you should run and scale your processes on Tallyfy :) Check out https://tallyfy.com/differe... - note that adding or changing people and their roles is not changing processes.

Re: Building With Workers KV, a Fast Distributed Key-Value Store

Amit Kothari — Sun, 20 Jan 2019 13:39:20 -0000

This is a great point @Ben Hadden - I can't count how many times the statement "let's store some data" turns into - well obviously we can filter/search/sort it now - right?! Something as simple as an online google sheet can do that.

Re: Building With Workers KV, a Fast Distributed Key-Value Store

Amit Kothari — Thu, 15 Nov 2018 09:12:38 -0000

This is worrying and very difficult for us to contemplate - since 10 seconds is an eternity for any real-life app.

"Eventually consistent, global consistency within 10 seconds"

With our specific application, which is real-time workflow tracking at https://tallyfy.com - we'd only consider moving to this (very promising) workers + KV combo if you could also embrace the real-time/socket-based use case. It's critical to both make and GET updates near-instantly for this to be truly useful and not laggy for a user experience on the web.

When/how do you intend to deal with real-time sync as a service between global, connected clients on a socket/channel ?

To summarize, I'd love Cloudflare to enable KV store + JS workers + real-time sockets (all at edge locations).

Re: Introducing Auth0 Extend: The new way to extend your SaaS

Amit Kothari — Fri, 28 Sep 2018 11:57:50 -0000

Thanks for this - it provides a great overview to a SaaS like us, on preventing feature bloat while also provide our customers with custom abilities to listen to events and write custom code.

For context - we run a workflow SaaS called https://tallyfy.com - and are evaluating this vs AWS Lambda functions to provide a platform to build custom features for our customers.

Could you elaborate on the precise difference between AWS Lambda functions and Auth0 Extend functions?

Re: Why Product Qualified Leads Are Rapidly Being Adopted in SaaS

Amit Kothari — Sat, 18 Nov 2017 16:38:35 -0000

This is a great post, and an approach that we are leveraging as a SaaS.

Beyond product-qualified leads, I had an idea for a future post that might be on a different subject.

Our vision is to have negative churn i.e. to move beyond "this guy is using the product so let's call him" and actually build self-serviced methods of stickiness and growth within our product in these ways:

- If you invite people into an app, it's less likely they will leave - since they have to explain things to co-workers
- If you start re-housing valuable data (in our case - business processes) - it's a sticky, new place for data that already existed
- If you become a surface for opinions and decisions (not just "collaboration") - the old way moves to our product, the new way.
- If you find a natural "hook" that depends on vanity (as evidenced by Facebook) - there's a higher stickiness long-term
- If you enable integrations, the pain of unwrangling them makes your SaaS more sticky.
- If you look to capture hard info (not random chat) - this makes your app an authoritative source.
- If you enable channel partners to educate customers (nobody talks about this?! - why??) - you build a great, long-term barrier to entry. In a SaaS like ours, we are uniquely able to use channels, even though direct signup is favoured.

... and some other secret things we're hacking into Tallyfy in the next few months :)

Amit
https://tallyfy.com

Re: How to Build a Brand That Attracts A-Players

Amit Kothari — Sat, 18 Nov 2017 16:21:45 -0000

Great post Eric! We're trying to build an awesome recruitment page off the back of this advice - https://tallyfy.com/careers

In case anyone's wondering about making employee onboarding (point 6 above) more consistent and effective, our app - Tallyfy, is often used for precisely this purpose.

Re: Introducing Modern Approvals in Flow

Amit Kothari — Sun, 29 Oct 2017 21:01:24 -0000

For anyone requiring more advanced approval functionality and proper tasking, etc - https://tallyfy.com can look into building an integration for Microsoft Flow. Just let us know through our website.

Re: The Next Recession is Set to Reshape Our Economy. Here’s How.

Amit Kothari — Sat, 21 Oct 2017 19:17:47 -0000

I think the candy vs vitamin vs painkiller is what ultimately matters, in a recession scenario. Bodes well for https://tallyfy.com imo :)

Re: VIDEO: How BP3 Helps Customers With Digital Transformation

Amit Kothari — Sat, 30 Sep 2017 18:15:16 -0000

Nice post. I would also add the perspective around digital transformation being about people, not just systems/IT:

https://tallyfy.com/digital...

Re: Why Build a Productized Service (Instead of Software)?

Amit Kothari — Tue, 05 Sep 2017 16:22:35 -0000

This makes a ton of sense. The only way to run a productized service is to automate repeatable workflows between people - and Tallyfy is a perfect platform for it.

Re: Announcing Microsoft’s Coco Framework for enterprise blockchain networks

Amit Kothari — Thu, 24 Aug 2017 22:11:42 -0000

Mark, this is interesting. We run an app (Tallyfy) that lets anyone build a step by step business process - with conditional branching and permissioning (i.e. who does which step) as well as form field capture of data from each participant.

At some point, Tallyfy will be a smart-contract-maker-for-your-grandma - the making of a smart contract would be simplified. We hope to be able to effectively "spin up" a collaborative Coco blockchain for a given process (once agreed) so that the running of the actual process between distributed partners like a customer and a company synced to our app/API at https://tallyfy.com - which just happens to be "another node" in the cluster.

Basically, the actual point and click "making" of a smart contract is what we aim to solve with Tallyfy. Once defined, running it would involve one-click "deploying" of that process it on an (Azure-based?) Coco cluster.

Now the interesting part is how a constantly evolving smart contract is handled. Imagine a process between people that frequently improves or changes. Is there any way of reaching consensus amongst all nodes for an updated version of a business process?

One requirement that a process tool like ours deals with - is the idea of a sequential or parallel process. Imagine a step in a contract that must be done before other steps, or certain steps that can be done by a given participant at any time. Can those enforcement rules also be built into smart contracts?

In business processes - you have steps that don't show up unless some other step has a specific input e.g. "Yes/No". I presume that would be embedded in a smart contract too?

Finishing Coco is then, almost like finishing the periodic table - where you know some elements, but you also know which ones are missing.

The completeness of Coco can be measured precisely by the ability to support known characteristics of business processes between different people, so that's what we're watching for at Tallyfy.

Look forward to 2018 and this being open sourced :)

Re: Making the Case for Third Party Office 365 Integrations

Amit Kothari — Sun, 20 Aug 2017 12:32:15 -0000

William - I'm afraid I disagree. "Supporting 3rd party integration" breaks down when you actually look into building a good integration. I can tell you this from actual experience with the latest API's for Outlook.

On your second point - "minimal ecosystem" - that's the point I was making. If you can't get enough third-parties interested, for reasons that are clear to third-parties, customers should be very worried. You're now stuck with your poor decision to implement Sharepoint or go in deep on the Microsoft stack. Second rate add-ons and plugins don't impress users.

"The business" in this case is not IT. It's business users who want integration with everything. IT cannot respond with "we'll build it for you" or "you have to stick with Microsoft" since the whole point of moving to the Office 365 was reducing IT costs and the availability of integrations.

There is zero chance that a Microsoft stack actually delivers what business users want and need. That's why shadow IT exists - people will use and pay for apps with or without the support or budget of IT.

In fact, most IT spending is moving away from IT:

"The share of worldwide corporate IT spending that is funded by non-IT business units is forecast to reach 47% in 2019"
https://www.idc.com/getdoc....

... hence IT defending Microsoft as an "open" platform holds little weight with business users.

Re: Making the Case for Third Party Office 365 Integrations

Amit Kothari — Sun, 20 Aug 2017 11:18:20 -0000

This would be Office.js in relation to cloud Outlook only. I'm not sure I understand your point though. The existence of API's doesn't solve business problems.

The real idea of this article is that third-party plugins are essential so that you don't have to hire armies of IT people to build and maintain custom code - which is the biggest problem faced by IT today.

The point I was trying to make is that the ecosystem around O365 is pretty weak, especially when compared to Google's. In addition, as mentioned above - trying to create an integration looks great on the marketing page, until you actually try to do it.

Re: Making the Case for Third Party Office 365 Integrations

Amit Kothari — Sat, 19 Aug 2017 23:08:55 -0000

Like with many things Microsoft, the devil is in the details. Many API methods are missing crucial capabilities and building an integration is far more complicated than it initially looks. If you search for actual developer traction on say Google Suite vs Office 365 on StackOverflow - the difference is obvious.

In any case - we have managed to build a strong Outlook integration that transforms forms, approvals and workflows for all internal business users of Outlook on Office 365. If you are interesting in learning more about that - please get in touch via our website, since it's currently under wraps to the public.

Re: The Best Place to Startup is at Home

Amit Kothari — Fri, 07 Apr 2017 09:52:15 -0000

Wade, we're based in St. Louis, Missouri - not far from you. We felt this problem so acutely that I actually asked a Wall Street Journal reporter to write a story about it:

https://www.wsj.com/article...

We've made a platform that lets you track and improve any human-driven workflow. Zapier is a natural partner for the automated stuff :)

We also want to invent a "human in the loop" zap which helps data science teams/devs do human-driven tasks which then feed back into ML models to continuously improve them.

Happy to chat about the whole Missouri/Silicon Valley thing :)

https://tallyfy.com/about-t...

Re: How to Automatically Track Your Blog's Traffic with Google Sheets, Google Analytics, and Zapier

Amit Kothari — Wed, 05 Apr 2017 12:20:56 -0000

Wow this is awesome - great advice. We do track all our posts, and content marketing plan on a Google Spreadsheet, so this totally makes sense. However, we do 35 QA checks via Tallyfy for content marketing - which is more of the elaborate QA as opposed to the basic data like views, etc.

In addition, I want to work out a way of doing this for any given URL, as opposed to just the latest URL's in the RSS feed - is there any obvious way to achieve that? For example, it it possible to write a Google Sheets function that runs a Zap or something like that?

That way - you could simply write the code as = MYFUNCTION (ZAPID, something) on the Google Sheet itself. That would be awesome!

At https://tallyfy.com - one of our other issues is to get TOTAL views for a given post - so I think I will modify your Google Analytics part above to obtain this.

The other thing I would love to know is how to pull in search volume and CPC for a given search term/target into a Google Sheet. I guess a Zap that obtains this from the Google Adwords API would be awesome - since we often ask "is it worth writing this post?" before we actually write it.

Re: BPM BUZZ – From BPM for Schools to Change Management

Amit Kothari — Tue, 04 Apr 2017 09:23:05 -0000

Thanks for the mention :)

- Amit
CEO, https://tallyfy.com