Let's discuss on Java

02/05/2026

“We’re not building features…
We’re reverse engineering the business.”
Old Business Analyst left the company.
Took all the domain knowledge with them.
New BA joins.
Opens Jira.
Writes:
“As discussed, implement the changes.”
No discussion. No context. No clarity.
Now the real work begins…
Read 10-year-old legacy code
Decode magic numbers
Guess business rules
Validate with production behavior 😄
At this point…
We’re not developers anymore.
We’re archaeologists.
Digging through code
Trying to understand what the business meant.

“In brownfield systems…
Code is not just implementation.
It’s undocumented business logic.”

This is why:
Ubiquitous language matters
Documentation matters
Domain knowledge transfer matters
Otherwise…
Every new feature becomes a discovery mission

01/05/2026

“We didn’t have a system bug…
We had a vocabulary bug.” 👇

ubiquitous language (DDD)

28/04/2026

Business VS NFR
“In tech, just like cricket…
We celebrate what’s visible.”
“Features get applause.
Architecture prevents collapse.”
“The irony?
The better the architect…
The less you notice them.”

27/04/2026

if you know you know

26/04/2026

“The Blame Game”

25/04/2026

“Client asked for a simple app…
I gave them microservices, Kafka, and trust issues.”

21/04/2026

Your distributed system isn’t slow.
It’s undefined!!!
No latency budget.
No failure model.
No consistency strategy.
But yeah… “it scales” 😏
💣 5 NFRs you can’t fake:
Latency → Do you even know your P99?
Availability → What breaks if 1 service dies?
Consistency → Strong or eventual… where?
Fault tolerance → Retry or just pray?
Observability → Can you debug at 2 AM?
⚡ Brutal truth:
Slow system?
👉 Chatty calls + no latency budget
System down?
👉 No circuit breaker + no isolation
🧠 Real architects don’t start with tech.
They start with:
👉 Constraints → then design
⚔️ Debate this:
Do you have P99 defined per API?
Can your system survive 1 node failure right now?
Or is it just “working in dev”?
Drop your architecture 👇
Let’s see if it’s designed… or just deployed.
Add to my circle, to get more insight
https://chat.whatsapp.com/DSXtWpvkTFPFlVaBezT5Pj?mode=gi_t
​

20/04/2026

Your “Event-Driven Architecture” is probably just glorified async CRUD.
⚡ 4 patterns as per Martin Fowler.
Pick wrong → system suffers silently.
Event Notification
Just a signal. Nothing more.
👉 Use when: You only need a trigger, not data.
Event-Carried StateEvent = full context. No callbacks.
👉 Use when: You hate chatty services.
Event SourcingState = replay of events.
👉 Use when: Audit, history, correctness matter.
CQRSRead ≠ Write.
Split them.
👉 Use when:
Reads are killing your system.
💣 Truth
nobody tells you:
Most teams don’t “design” EDA…They just add Kafka and pray.
🧠 Reality check:
Are your services still calling each other after events?
Is your “EDA” just REST + messaging glued together?
CQRS avoided… because it exposes bad design?
So tell me
👇Which pattern are you actually using…and which one are you avoiding on purpose?
Add to my circle, to get more insight
https://chat.whatsapp.com/DSXtWpvkTFPFlVaBezT5Pj?mode=gi_t
​

19/04/2026

Partition tolerance is not a choice.It’s a failure you’re already living with.

Your system is running fine.Then suddenly:
• Network latency spikes
• Requests start timing out
• Some services can’t reach othersWelcome to a network partition.

CAP Theorem: Now you have only 2 real choices:

1. Stay Available (AP)→ Serve requests with possibly stale data
2. Stay Consistent (CP)→ Reject requests until the system stabilizesYou don’t get both.
--------------------
💣 Real Production ScenarioUser tries to transfer money:
• Service A debits ₹1000
• Service B (credit) is unreachableNow what option you choose?
Option 1 (Availability):
Show success → retry later👉 Risk: double credit / inconsistency
Option 2 (Consistency):Fail the transaction👉 User sees failure, but system stays correct
⚠️ Hard Truth
Most systems don’t “choose” CAP.They accidentally drift into it during failure.
🧠 Architect’s Decision
Before production, you must answer:
• Is stale data acceptable here?
• Can I retry safely (idempotency)?
• What’s worse: wrong data or no response?
Partition tolerance isn’t about handling failure.
It’s about deciding your failure behavior in advance.
For a payment system:
Would you choose Availability or Consistency during a partition?
No theory.
Answer like you’re in production.
Add to my circle, to get more insight

https://chat.whatsapp.com/DSXtWpvkTFPFlVaBezT5Pj?mode=gi_t
​

01/06/2025

s
🧘‍♂️🐼 What happens when Po needs to pick the right move during battle – at runtime?
Shifu doesn’t give him a fixed answer.
He teaches him the Strategy Pattern — a way to swap behaviors dynamically, without rewriting his warrior code. 🧘‍♂️⚔️
In this new blog, Po discovers how to:
✅ Isolate behaviors into interchangeable strategies
✅ Select algorithms at runtime like a true coding warrior
✅ Replace if-else chaos with clean object-oriented design
✅ Implement Strategy Pattern with real-world examples in Java
✅ Even refactor it using Lambdas for modern clarity!
📜 Whether it's dodging, punching, or logging—Po learns that choosing the right strategy is the real kung fu.
🚀 Perfect for:
🔹 Java developers
🔹 Interview prep
🔹 Clean code warriors
👉 https://javaonfly.blogspot.com/.../kung-fu-code-master...

Call now to connect with business.